Role of Body Fat and Metabolic Rate in Site-Specific Fracture Risk: A 20-Year Taiwanese Cohort Study

Introduction

LIMITATIONS OF BODY MASS INDEX IN FRACTURE RISK PREDICTION:

Body mass index (BMI) has been conventionally used as a surrogate marker for fracture risk, with studies suggesting that both low and high BMI can influence skeletal fragility through different mechanisms [9–12]. While higher BMI is generally associated with increased BMD due to mechanical loading effects, it does not always confer protection against fractures [11]. In obese individuals, greater soft tissue padding may reduce impact forces during falls, yet obesity is also linked to poorer bone microarchitecture and increased risk of certain fractures [12]. Conversely, low BMI is strongly associated with osteoporosis and reduced cortical bone thickness, predisposing individuals to fragility fractures [13].

Beyond BMI, body composition parameters such as body fat percentage (BFP) and basal metabolic rate (BMR) have emerged as critical determinants of skeletal integrity [14,15]. BFP influences bone turnover and adipokine-mediated bone remodeling, whereas BMR reflects energy metabolism, which is crucial for maintaining bone homeostasis [16–18]. Recent findings indicate that lean mass exerts a more substantial effect than fat mass on BMD and fracture resistance, further questioning the reliability of BMI-based assessments [19–21].

ETHNIC VARIABILITY IN FRACTURE RISK AND THE NEED FOR POPULATION-SPECIFIC STUDIES:

Ethnic differences in bone density, body composition, and fracture susceptibility necessitate population-specific risk assessments. Studies comparing Asian and Western populations have shown that Asians tend to have lower BMI but higher BFP at comparable levels of skeletal strength [22–25]. Such disparities may lead to underestimation of fracture risk in Asian populations when using BMI-centric screening models. Therefore, it is imperative to refine fracture risk prediction by integrating BFP, BMR, and BMD assessments to develop a more comprehensive and ethnicity-specific risk stratification model.

THE IHI DATA PLATFORM: A HIGH-RESOLUTION MEDICAL DATABASE:

To address these gaps in fracture risk prediction, this study utilized the iHi Data Platform of China Medical University Hospital (CMUH), a multi-dimensional, AI-enhanced research database that integrates 19 years of electronic medical records (EMRs), dual-energy X-ray absorptiometry (DXA)-measured BMD data, and metabolic health indices [26]. The iHi Data Platform is the largest phenome-genome-environmental data repository in Asia, encompassing de-identified clinical, genetic, and environmental exposure data from over 3 million patients [26]. The platform’s ISO-certified, systematic data cleaning protocols and modular architecture (Data LEGO framework) ensure high-quality, reproducible clinical datasets for advanced epidemiological analyses [26].

A key advantage of the iHi Data Platform is its capability to integrate diverse patient-level data, including BMI, BFP, BMR, and BMD, providing a comprehensive assessment of metabolic and skeletal health. Previous studies utilizing this platform have generated high-resolution clinical insights into metabolic disorders and chronic diseases, reinforcing its reliability as a robust research tool [27,28]. Given the growing importance of big data in precision medicine, leveraging the iHi Platform enables more accurate fracture risk stratification and data-driven improvements in osteoporosis management.

STUDY OBJECTIVES:

Considering the limitations of BMI as a sole predictor of fracture risk, this study aimed to examine the interplay between BMI, BFP, BMR, and BMD in determining site-specific fracture risk. Specifically, this retrospective cohort study utilized the iHi Data Platform of CMUH (2000–2020) to evaluate the associations between BMI, BMD, BFP, BMR, and fracture risk in a cohort of 9583 Taiwanese individuals. The findings from this study are expected to contribute to refining current osteoporosis screening guidelines and improving personalized risk assessment models.

Material and Methods

ETHICS STATEMENT:

This study was conducted with approval from the Institutional Review Board of China Medical University Hospital (CMUH) in Taiwan (approval no. CMUH112-REC1-026). Access to the integrated Healthcare information (iHi) Data Platform of CMUH was granted following institutional protocols for research purposes, ensuring compliance with data security and privacy regulations. Given the retrospective and de-identified nature of the data, the requirement of informed consent from individual patients was waived by the ethics committee, in accordance with local regulations and international standards, including the Declaration of Helsinki.

STUDY DESIGN AND DATA SOURCE:

This investigation employed a retrospective cohort design utilizing the iHi Data Platform of CMUH. The iHi Platform is a comprehensive healthcare data ecosystem that integrates electronic medical records (EMRs), imaging data, laboratory results, and medication histories from over 3 million patients into a structured database spanning a 19-year period (2000–2020). This platform uses automated validation pipelines to ensure data accuracy, integration, and de-identification, transforming raw clinical data into a resource for evidence-based research [26].

The platform’s reliability and validity are substantiated by its use in numerous scientific publications and its compliance with international standards for medical data de-identification, including ISO 29100 and 29191 and CNS 29100-2 certifications [26]. For this study, we specifically extracted dual-energy X-ray absorptiometry (DXA) records for bone mineral density (BMD) assessment, along with associated clinical data, to investigate relationships between body composition parameters, BMD, and fracture risk.

STUDY POPULATION:

Our study included adults aged 18 years or older who underwent DXA scans at CMUH between 2000 and 2020 (Figure 1). To ensure data completeness, we required participants to have recorded body composition metrics, including body mass index (BMI), body fat percentage (BFP), and basal metabolic rate (BMR), within 6 months of their DXA assessment. Additionally, only individuals with longitudinal medical records allowing for adequate follow-up to track fracture outcomes were included.

We excluded patients with fractures caused by acute trauma (such as motor vehicle accidents or falls from height) within 2 weeks before the recorded fracture event, as these are traumatic rather than osteoporotic fractures. Patients with a history of prolonged corticosteroid use exceeding 3 consecutive months prior to fracture were also excluded due to the well-documented impact of corticosteroids on bone metabolism. Furthermore, individuals with incomplete DXA records or missing key clinical variables (BMI, BFP, or BMR) were omitted to maintain data integrity and reliability.

OUTCOME DEFINITION AND FRACTURE CLASSIFICATION:

The primary outcome of this study was the occurrence of site-specific fractures, which were categorized into several groups for comprehensive analysis. Overall fracture was defined as any fracture recorded within a participant’s medical history. Vertebral fractures specifically referred to those of the spine, identified using International Classification of Diseases, 9^th Revision, Clinical Modification (ICD-9-CM) code 805.xx and ICD-10-CM codes S22.0-S22.1 and S32.0. Non-vertebral fractures occurred at sites other than the spine and were identified using ICD-9-CM codes 812.xx and 813.xx, and ICD-10-CM codes S42.xx and S52.xx.

Hip fractures were identified using ICD-9-CM code 820.xx and ICD-10-CM codes S72.0-S72.2. Upper limb fractures included those of the arms and were identified using ICD-9-CM codes 810.xx through 819.xx and ICD-10-CM codes S42.xx and S52.xx. Lower limb fractures encompassed those of the legs, identified using ICD-9-CM codes 821.xx through 829.xx and ICD-10-CM codes S72.3 through S82.0.

To ensure accurate classification, fracture events were confirmed through a comprehensive review of radiology reports from CMUH imaging records, EMR diagnosis codes, and prescription histories related to fracture treatments, such as bisphosphonates, calcium, and vitamin D supplements. This multi-faceted verification approach enhanced the reliability of our fracture outcome assessment.

EXPOSURE VARIABLES: BODY COMPOSITION AND BMD MEASUREMENTS:

The primary exposure variables in this study focused on body composition and bone mineral density. BMD was assessed using DXA scans targeting the lumbar spine (L1–L4), total hip, and femoral neck. Based on the World Health Organization criteria, participants were classified according to their T-scores: normal BMD (T-score ≥−1.0), osteopenia (−2.5< T-score <−1.0), or osteoporosis (T-score ≤−2.5).

BMI was calculated as weight in kilograms divided by height in meters squared (kg/m²). Following Taiwan-specific BMI classification guidelines, participants were categorized as underweight (<18.5 kg/m²), normal weight (18.5–24 kg/m²), overweight (24–27 kg/m²), or obese (≥27 kg/m²). Weight and height measurements closest to the BMD assessment date were used for these calculations.

BFP and BMR were derived from either bioelectrical impedance analysis (BIA) or DXA-derived equations. BFP was categorized according to sex-specific thresholds: low (<25% in men, <32% in women), normal (25–35% in men, 32–42% in women), or high (>35% in men, >42% in women). BMR was calculated using the Mifflin-St Jeor equation, with adjustments for age, sex, and body weight, and was categorized as <1000, 1000–1500, or ≥1500 for analysis purposes.

Relevant comorbidities including diabetes mellitus, psychoactive substance use, obesity, rheumatoid arthritis, nutritional deficiencies, and hyperparathyroidism were identified using ICD-9-CM, ICD-10-CM, and Anatomical Therapeutic Chemical (ATC) codes. These conditions were included as potential confounders in our analyses given their known associations with bone health and fracture risk.

STATISTICAL ANALYSIS:

All statistical analyses were performed using R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are presented as mean±standard deviation (SD) or median (interquartile range), while categorical variables are expressed as count (percentages). Group differences were assessed using analysis of variance (ANOVA) or Kruskal-Wallis tests for continuous variables and chi-square tests for categorical variables.

To analyze time-to-fracture outcomes, we constructed Kaplan-Meier survival curves, with log-rank tests comparing fracture-free survival across BMI, BMD, and BFP categories. Cox proportional hazards regression models were applied to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) for fracture risk. These models were adjusted for potential confounders, including age, sex, smoking status, comorbidities (diabetes, rheumatoid arthritis, hyperparathyroidism, nutritional deficiencies), medication use (corticosteroids, osteoporosis treatments), and body composition parameters (BMI, BFP, BMR).

To explore whether BFP and BMR mediated the relationship between BMI and fracture risk, we conducted a mediation analysis using the ‘mma’ R package. This analysis employed a causal mediation framework with bootstrap resampling (5000 iterations) to estimate mediation effects. The total effect represented the overall association between BMI and fracture risk, encompassing both direct and indirect pathways. The direct effect quantified the portion of this association independent of BFP and BMR, while the indirect effect measured the extent to which the relationship was mediated through these factors.

Several sensitivity analyses were performed to assess the robustness of our findings. Multiple imputation using the multivariate normal imputation method was employed for missing data. Stratified analyses by sex and age groups (≤65 years vs >65 years) were conducted to identify potential demographic variations in associations. An additional sensitivity analysis excluding corticosteroid users (n=1212) was performed to evaluate potential medication-related biases in the primary findings. Statistical significance was established at p<0.05 for all analyses.

Results

STUDY POPULATION CHARACTERISTICS:

A total of 9583 participants who underwent dual-energy X-ray absorptiometry (DXA) scans at China Medical University Hospital (CMUH) between 2000 and 2020 were included in this study. The demographic and clinical characteristics of the study population are summarized in Table 1. Most participants were female (78.5%, n=7524), and the mean age was 59.5±13.1 years. The age distribution showed that most participants were 46–65 years old (55.2%, n=5,292) or older than 65 years (33.3%, n=3195), while younger participants comprised a smaller proportion (age ≤25 years: 0.9%, n=81; age 26–45 years: 10.6%, n=1015).

The mean BMI value was 23.8±3.8 kg/m². Most participants had a BMI classified as normal to overweight (24–27 kg/m², 51.2%, n=4905) or obese (≥27 kg/m², 43.4%, n=4159), while only 5.4% (n=519) were underweight (<24 kg/m²). BMD assessment revealed that 26.0% (n=2490) of participants had osteoporosis (T-score ≤−2.5), 49.8% (n=4774) had osteopenia (−2.5< T-score <−1.0), and 24.2% (n=2319) had normal BMD (T-score ≥−1.0). The mean BMD value was 0.87±0.15 g/cm².

BFP measurements indicated that most participants (69.2%, n=6629) had above-normal values, 30.6% (n=2929) had normal values, and only 0.3% (n=25) had below-normal values. The mean BFP was 34.5±7.1%. For basal metabolic rate (BMR), most participants (72.9%, n=6981) were within the 1000–1500 range, while 19.4% (n=1860) had BMR <1000, and 7.7% (n=742) had BMR ≥1500. The mean BMR was 1161.6±203.8.

The prevalence of comorbidities was highest for diabetes mellitus (12.4%, n=1187), followed by rheumatoid arthritis (4.1%, n=389), hyperparathyroidism (0.4%, n=37), psychoactive substance use (0.3%, n=24), nutritional deficiencies (0.1%, n=10), and obesity (0.1%, n=5).

FRACTURE INCIDENCE AND SITE-SPECIFIC DISTRIBUTION:

During a median follow-up of 8.7 years (interquartile range: 5.3–12.2 years), a total of 1672 fractures (17.4% of the cohort) were recorded. The site-specific distribution revealed that vertebral fractures occurred in 633 cases (6.6%), non-vertebral fractures in 1039 cases (10.8%), hip fractures in 302 cases (3.1%), upper limb fractures in 425 cases (4.4%), and lower limb fractures in 512 cases (5.3%) (Table 2).

The incidence rate of fractures was significantly higher in females (1.79 per 100 person-years) compared to males (1.14 per 100 person-years). Age was strongly associated with fracture risk, with participants older than 65 years showing the highest incidence rate (3.74 per 100 person-years) compared to those aged 46–65 years (1.16 per 100 person-years), 26–45 years (0.41 per 100 person-years), and ≤25 years (0.76 per 100 person-years).

Fracture incidence varied substantially across BMD categories, with the highest rates observed in participants with osteoporosis (3.75 per 100 person-years), followed by those with osteopenia (1.34 per 100 person-years) and normal BMD (0.63 per 100 person-years). Similarly, participants with below-normal BFP had higher fracture incidence (2.36 per 100 person-years) compared to those with normal BFP (0.83 per 100 person-years), while those with above-normal BFP had an increased incidence (2.11 per 100 person-years). Lower BMR was associated with higher fracture incidence, with rates of 3.61, 1.39, and 0.59 per 100 person-years for BMR categories <1000, 1000–1500, and ≥1500, respectively.

BONE MINERAL DENSITY AS A PREDICTOR OF FRACTURE RISK:

In Cox proportional hazards models adjusted for age, sex, BMI, comorbidities, and medication use, BMD emerged as a strong independent predictor of fracture risk across all skeletal sites. Compared to individuals with osteoporosis (T-score ≤−2.5), those with osteopenia had a 49% lower adjusted fracture risk (adjusted hazard ratio [aHR]=0.51, 95% confidence interval [CI]: 0.44–0.58, p<0.001). The protective effect was even more pronounced in participants with normal BMD, who demonstrated an 82% lower adjusted fracture risk (aHR=0.18, 95% CI: 0.15–0.22, p<0.001) compared to those with osteoporosis.

The relationship between BMD and fracture risk showed site-specific variations (p for interaction <0.01). The strongest association was observed for vertebral fractures, where normal BMD conferred an 82% reduction in risk (aHR=0.18, 95% CI: 0.12–0.27, p<0.001) compared to osteoporosis. For non-vertebral fractures, normal BMD was associated with a 58% reduction in risk (aHR=0.42, 95% CI: 0.32–0.55, p<0.001). Upper and lower limb fractures showed similar protective associations with higher BMD, with adjusted hazard ratios of 0.48 (95% CI: 0.32–0.71, p<0.001) and 0.38 (95% CI: 0.25–0.56, p<0.001), respectively, for normal versus osteoporotic BMD.

BODY MASS INDEX AND FRACTURE RISK: THE BMI PARADOX:

The relationship between BMI and fracture risk demonstrated a complex, non-linear pattern. Compared to individuals with normal BMI (18.5–24 kg/m²), underweight individuals (BMI <18.5 kg/m²) had a significantly higher risk of fractures (aHR=1.75, 95% CI: 1.29–2.36, p<0.001). However, contrary to expectations, overweight status (BMI 24–27 kg/m²) conferred only a modest, non-significant reduction in fracture risk (aHR=0.92, 95% CI: 0.77–1.08, p=0.12). Surprisingly, obese individuals (BMI ≥27 kg/m²) did not experience significant protection from fractures (aHR=1.04, 95% CI: 0.89–1.21, p=0.58).

Stratified analysis by sex revealed stronger associations in females (p for sex interaction=0.017), with higher fracture risk in underweight women but no significant protective effect of obesity. The fracture incidence rate was 1.79 per 100 person-years in females compared to 1.14 per 100 person-years in males, resulting in a crude hazard ratio of 0.65 (95% CI: 0.55–0.77, p<0.001) for males versus females. However, after adjustment for covariates, the sex difference became non-significant (aHR=0.87, 95% CI: 0.72–1.06, p=0.164).

The site-specific analysis revealed heterogeneous associations between BMI and fracture risk across different skeletal regions. For vertebral fractures, neither overweight (aHR=0.90, 95% CI: 0.62–1.30, p=0.574) nor obese status (aHR=1.16, 95% CI: 0.77–1.75, p=0.477) showed significant associations compared to normal BMI. Similar non-significant patterns were observed for non-vertebral, hip, upper limb, and lower limb fractures, challenging the conventional wisdom that higher BMI uniformly protects against fractures.

MEDIATING ROLE OF BODY FAT PERCENTAGE AND BASAL METABOLIC RATE IN FRACTURE RISK:

To explore the mechanisms underlying the complex BMI-fracture relationship, we conducted a mediation analysis examining the roles of BFP and BMR as potential mediators. The analysis revealed significant associations between these factors and both BMI and fracture risk, with important sex-specific differences.

In females, BMI was significantly associated with BFP (p=0.001), BMR (p<0.001), BMD (p<0.001), diabetes mellitus (p=0.01), psychoactive substance use (p=0.042), and rheumatoid arthritis (p=0.049). Fracture risk was significantly associated with BFP (p<0.001), BMR (p<0.001), BMD (p<0.001), and diabetes mellitus (p=0.032). The relative mediating effect of BFP on the BMI-fracture relationship was substantial, accounting for 50% of the total effect (Figure 2, Table 3).

In males, BMI was significantly associated with BMD (p<0.001), while fracture risk was significantly associated with BFP (p<0.001), BMR (p<0.001), BMD (p<0.001), and diabetes mellitus (p=0.008). The relative mediating effect of BFP on the BMI-fracture relationship was even more pronounced in males, explaining 79% of the total effect (Figure 3, Table 3).

When examining the direct associations between these mediators and fracture risk, we found that normal BFP significantly reduced overall fracture risk by 77% compared to low BFP (aHR=0.23, 95% CI: 0.07–0.72, p=0.012). Above-normal BFP also showed a protective effect (aHR=0.27, 95% CI: 0.08–0.87, p=0.028), but with no additional benefit beyond normal BFP. For vertebral fractures specifically, normal BFP was associated with an 85% reduction in risk compared to low BFP (aHR=0.15, 95% CI: 0.02–0.97, p=0.047).

Higher BMR was consistently associated with lower fracture risk, with adjusted hazard ratios of 0.78 (95% CI: 0.66–0.91, p=0.002) and 0.54 (95% CI: 0.36–0.82, p=0.004) for BMR categories 1000–1500 and ≥1500, respectively, compared to BMR <1000. This protective effect was particularly pronounced for vertebral fractures, with corresponding adjusted hazard ratios of 0.62 (95% CI: 0.49–0.78, p<0.001) and 0.30 (95% CI: 0.14–0.63, p=0.002).

COMORBIDITIES AND FRACTURE RISK:

Among the examined comorbidities, diabetes mellitus emerged as a significant risk factor for fractures (Table 4). Participants with diabetes had a 48% higher adjusted fracture risk compared to those without (aHR=1.48, 95% CI: 1.26–1.74, p<0.001). Psychoactive substance use was associated with nearly 3 times higher fracture risk (aHR=2.95, 95% CI: 1.09–7.98, p=0.033). Other comorbidities, including obesity (aHR=4.26, 95% CI: 0.60–30.42, p=0.149), nutritional deficiencies (aHR=2.62, 95% CI: 0.63–10.86, p=0.184), rheumatoid arthritis (aHR=1.01, 95% CI: 0.73–1.41, p=0.943), and hyperparathyroidism (aHR=0.78, 95% CI: 0.25–2.44, p=0.675), showed non-significant associations with fracture risk after adjustment for confounders.

Mediation analysis indicated that diabetes mellitus partially mediated the relationship between BMI and fracture risk in females, but not in males, highlighting the sex-specific nature of these associations.

SENSITIVITY ANALYSES:

Multiple sensitivity analyses were conducted to assess the robustness of our findings. Multiple imputation for missing data confirmed the consistency of our results, supporting their validity. Stratified analysis by sex and age demonstrated that older adults (>65 years) with low BFP had the highest fracture risk (p<0.001), while women with high BFP did not exhibit additional fracture protection beyond that conferred by normal BFP (p=0.21).

To address potential confounding by medication use, we conducted a sensitivity analysis excluding corticosteroid users (n=1212), which yielded similar hazard ratios to the primary analysis, suggesting that glucocorticoid exposure did not significantly bias our results. Additional analyses stratified by menopausal status in women confirmed the persistence of the observed patterns across pre- and post-menopausal groups, although the associations were stronger in post-menopausal women.

Discussion

SUMMARY OF KEY FINDINGS:

This study investigated the association between body mass index (BMI), body fat percentage (BFP), basal metabolic rate (BMR), and bone mineral density (BMD) in predicting site-specific fracture risk. The findings demonstrated that while higher BMI was generally protective against vertebral fractures, it did not significantly reduce the risk of hip or lower limb fractures. Instead, BFP and BMR exhibited independent and contrasting effects, with higher BFP being a risk factor and higher BMR being protective. These results indicate that fracture risk assessment should not rely solely on BMI and highlight the importance of considering body composition variables in clinical evaluations.

The survival analysis confirmed that osteoporosis significantly reduces fracture-free survival, particularly among individuals with low BMR and high BFP. Mediation analysis revealed that BFP and BMR partially mediated the association between BMI and fracture risk, reinforcing the complex metabolic and biomechanical interactions influencing skeletal fragility. These findings challenge the traditional reliance on BMI as a primary predictor of fracture risk and support a more comprehensive, integrated approach that includes BFP, BMR, and BMD measurements.

COMPARISON WITH PREVIOUS STUDIES:

The relationship between BMI and fracture risk has been extensively investigated, but prior studies have yielded conflicting results, often due to ethnic differences, variations in body composition assessment, and the influence of confounding variables [29–32]. Several large-scale cohort studies have reported that higher BMI is associated with reduced fracture risk, primarily due to the mechanical loading effect on bone [9,10,33]. However, our study aligns with recent findings that emphasize the heterogeneous impact of body composition, particularly the differential effects of lean mass and adiposity on skeletal health.

In contrast to earlier research that suggested higher BMI confers universal protection against fractures [9,29,31], this study demonstrates that this protective effect is not uniform across different fracture sites. Prior studies have reported that higher adiposity may exert both beneficial and detrimental effects on bone. On one hand, greater soft tissue padding can reduce impact forces during falls, thereby protecting against hip and vertebral fractures [34]. On the other hand, excessive adiposity is associated with metabolic alterations that negatively influence bone remodeling, leading to increased skeletal fragility, particularly in the lower limbs [11,35]. These findings are consistent with recent investigations highlighting that BFP is an independent predictor of fracture risk, with excess adiposity contributing to bone fragility through inflammatory cytokines and altered bone metabolism.

The role of BMR in fracture risk has been less frequently explored in previous studies. However, our findings suggest that higher BMR is protective against fractures, likely due to its association with greater lean mass and improved bone turnover. A growing body of evidence indicates that muscle mass and physical activity levels significantly contribute to bone strength, further supporting that energy metabolism plays a crucial role in maintaining skeletal integrity [36–38]. This observation underscores the limitations of BMI as a surrogate for fracture risk assessment, as it fails to account for variations in body composition and metabolic function.

CLINICAL IMPLICATIONS:

These findings have significant clinical implications for osteoporosis screening and fracture risk assessment. The current reliance on BMI-based classification for osteoporosis risk stratification may lead to misclassification of individuals who have normal BMI but abnormal body composition. Traditional BMI categories do not differentiate between lean mass and adiposity, which are critical determinants of bone strength and mechanical loading capacity. The results of this study suggest that BFP and BMR should be incorporated into routine fracture risk assessment to improve predictive accuracy and enable more personalized intervention strategies.

The findings also emphasize the importance of targeted prevention strategies for individuals at higher fracture risk due to adverse body composition profiles. Individuals with high BFP but low lean mass may benefit from structured resistance training programs to enhance musculoskeletal function, whereas those with low BMR may require nutritional and metabolic interventions to support bone health. These insights could inform more effective public health strategies aimed at reducing osteoporosis-related fractures by shifting the focus from BMI-based screening to body composition-inclusive risk models.

LIMITATIONS AND METHODOLOGICAL CONSIDERATIONS:

Despite the strengths of using a large, high-resolution dataset from the iHi Data Platform, this study has several limitations. The retrospective study design introduced potential biases related to data completeness and residual confounding. Although efforts were made to adjust for key confounders such as age, sex, comorbidities, and medication use, unmeasured variables such as physical activity levels, dietary intake, and genetic predisposition could influence the observed associations.

Another limitation relates to the use of DXA-derived BFP measurements. While DXA is considered a criterion standard for body composition analysis, it may not fully capture regional fat distribution and visceral adiposity, which are known to have distinct metabolic effects on bone health. Additionally, BMR was estimated using the Mifflin-St Jeor equation, which, although widely validated, may not accurately reflect individual variations in metabolic rate due to lifestyle and hormonal factors. Future studies incorporating direct calorimetry or metabolic chamber assessments may provide a more precise evaluation of energy metabolism in relation to fracture risk.

The study population was derived from a single-center dataset in Taiwan, which may limit the generalizability of findings to other ethnic groups. Given that ethnic differences in bone structure, fat distribution, and metabolism exist, further validation in multi-ethnic cohorts is warranted. Additionally, the follow-up period, while sufficient for identifying fractures, does not account for lifelong skeletal changes. Future prospective studies with longitudinal BMD and body composition tracking will be crucial for understanding the causal mechanisms linking metabolic health and fracture risk.

Conclusions

This large-scale retrospective cohort study, using the comprehensive iHi Data Platform of China Medical University Hospital, provides compelling evidence that the relationship between body composition, bone mineral density (BMD), and fracture risk is more nuanced than previously understood. Our findings demonstrate that while BMD remains the strongest predictor of fracture risk, with normal BMD conferring an 82% lower risk compared to osteoporosis, it alone is insufficient for comprehensive fracture risk assessment.

The relationship between body mass index (BMI) and fracture susceptibility reveals a complex pattern that challenges conventional understanding – underweight status significantly increases fracture risk, but obesity does not confer the universal protection previously assumed. This “BMI paradox” is explained through our mediation analysis, which identified body fat percentage (BFP) and basal metabolic rate (BMR) as critical intermediaries in the BMI-fracture relationship, accounting for 50% and 32% of this association, respectively. Normal BFP significantly reduced fracture risk by 77% compared to low BFP, while higher BMR was consistently associated with lower fracture incidence, particularly for vertebral fractures.

These findings have important clinical implications for fracture risk assessment in Asian populations, where body composition profiles differ significantly from Western cohorts. The integration of BFP and BMR measurements alongside traditional BMD assessments could substantially enhance fracture prediction accuracy and enable more personalized prevention strategies. Healthcare providers should consider comprehensive body composition analysis rather than relying solely on BMI or BMD when evaluating fracture risk, particularly in patients with diabetes mellitus, who demonstrated a 48% higher fracture risk in our study.

Future fracture prevention efforts should move beyond BMD-centric approaches to incorporate body composition metrics, enabling more precise risk stratification and targeted interventions. This study establishes a foundation for developing ethnicity-specific fracture risk models that account for the complex interplay between adiposity, metabolic activity, and bone health in diverse populations.