Underestimation of fracture risk by conventional FRAX thresholds in axial spondyloarthritis: a proposal for disease-specific thresholds

Abstract

Background:

The Fracture Risk Assessment Tool (FRAX) is widely used as an intervention threshold for postmenopausal osteoporosis but has not been validated specifically for axial spondyloarthritis (axSpA).

Objectives:

To validate FRAX thresholds and predictive ability for fragility fractures in axSpA patients.

Design:

A cross-sectional study.

Methods:

Patients diagnosed with axSpA according to the 2009 ASAS criteria, aged 40–90 years, and treated at Phramongkutklao Hospital between December 2019 and October 2024 were consecutively included. Bone mineral density (BMD), trabecular bone score (TBS), and vertebral fracture assessment were measured using Dual-energy X-ray absorptiometry. The Thai reference database calculated FRAX probabilities for major osteoporotic fracture (MOF) and hip fracture (HF). The predictive performance of FRAX (with and without BMD) was evaluated against actual fracture occurrence, using FRAX thresholds for MOF and HF at 10% and 3%, respectively. Optimal cutoffs for FRAX against fractures were determined using receiver operating characteristic (ROC) curves and the Youden index.

Results:

Among 125 axSpA patients (73.6% male; mean age 50.9 ± 10.8 years; disease duration 9.7 ± 9.9 years), osteoporotic fractures were found in 24.0%, with vertebral fractures being most common. Patients with fractures had significantly higher disease activity (Bath Ankylosing Spondylitis Disease Activity Index 4.2 vs 2.2, p = 0.004; Ankylosing Spondylitis Disease Activity Score (ASDAS) 3.1 vs 1.7, p < 0.001) and higher median FRAX scores for MOF (4.9% vs 1.9% without BMD; 4.7% vs 2.4% with BMD) and HF (0.8% vs 0.2% without BMD; 0.9% vs 0.5% with BMD; all p < 0.05). Higher ASDAS scores were independently associated with increased fracture risk (odds ratio 2.35, 95% confidence interval 1.39–3.98). FRAX thresholds (MOF ⩾10%, HF ⩾3%) showed low sensitivity (MOF: 0% without BMD, 18.5% with BMD; HF: 23.3% without BMD, 22.2% with BMD) but high specificity (>84%). Optimal ROC-derived FRAX cutoffs without BMD (MOF ⩾4%, HF ⩾1.5%) improved predictive accuracy (MOF: sensitivity 40%, specificity 85.3%, area under the ROC curve (AUC) = 0.770; HF: sensitivity 46.7%, specificity 75.8%, AUC = 0.674).

Conclusion:

FRAX (without BMD) is a viable predictor of fractures in axSpA, yet its optimal intervention thresholds (⩾4% MOF and ⩾1.5% HF) fall substantially below standard thresholds for postmenopausal women. Adjusting clinical thresholds downward is warranted to ensure timely prevention and better reflect the specific axSpA risk profile.

Plain language summary

Current FRAX Thresholds May Underestimate Fracture Risk in Axial Spondyloarthritis: The Need for Disease-Specific Thresholds

Conventional FRAX thresholds significantly underestimate fracture risk in patients with axial spondyloarthritis (axSpA), as these benchmarks were primarily derived from postmenopausal populations. This study evaluated the diagnostic performance of FRAX in a cohort of 125 patients (74% male; mean age 50.9 years), 30 of whom had sustained fragility fractures. While fracture patients exhibited significantly higher median FRAX scores than those without fractures (p < 0.001), standard treatment thresholds demonstrated poor clinical utility. At the conventional FRAX-major osteoporotic fracture (FRAX-MOF) threshold of more than or equal to 10% (without BMD), sensitivity was 0.0%. Notably, no patient in the cohort reached the traditional FRAX-MOF of more than or equal to 20%. Receiver Operating Characteristic (ROC) analysis revealed that FRAX is a robust discriminative tool, with FRAX-MOF without BMD achieving an Area Under the Curve (AUC) of 0.770. The inclusion of bone mineral density (BMD) did not enhance performance. To improve the sensitivity of FRAX, we propose disease-specific thresholds of more than or equal to 4% for FRAX-MOF and 1.5% for FRAX-HF without BMD, respectively. These recalibrated cutoffs provided a superior balance of sensitivity and specificity (56.7% and 78.9% for MOF; 40.0% and 85.3% for HF). Adopting these lower thresholds offers a practical, cost-effective strategy to identify high-risk axSpA patients.

Keywords

axial spondyloarthritis bone mineral density fracture fracture risk assessment FRAX osteoporosis trabecular bone score vertebral fracture

Introduction

Axial spondyloarthritis (axSpA) is a chronic inflammatory condition primarily affecting the axial joints, such as the spine and sacroiliac joint, characterized by the formation of syndesmophytes and fusion of the spine and sacroiliac joint. Despite the development of new bone or syndesmophytes, the prevalence of osteoporosis and vertebral fractures remains high, ranging from 19% to 62% and 30% to 40%, respectively.^1–3 A multicenter study conducted in Thailand found that the prevalence of osteoporosis and degraded bone by trabecular bone score (TBS) in axSpA patients was 9.6% and 19.1%, respectively, with vertebral fractures occurring in 7.5% of patients.² Notably, osteoporosis may occur early in the disease course,⁴ underscoring the importance of awareness and early detection for timely intervention.

Screening for osteoporosis and vertebral fractures in axSpA patients is recommended,⁵ due to their high prevalence, which is often asymptomatic, the symptoms may be masked by chronic inflammatory back pain. Dual-energy X-ray absorptiometry (DXA) is the standard tool for assessing bone mineral density (BMD); however, its sensitivity for detecting vertebral fractures is limited, with only 10% for the lumbar spine (LS) BMD and 16.7% for the femoral neck (FN) BMD.² Furthermore, LS BMD may be overestimated in axSpA due to syndesmophyte formation, potentially leading to underestimation of fracture risk.

Fracture risk assessment tool (FRAX) is a web-based algorithm that calculates the 10-year probability of major osteoporotic fractures (FRAX-MOF) and hip fractures (FRAX-HF) based on various clinical factors, including age, gender, weight, height, comorbidities, history of alcohol consumption, smoking, corticosteroid usage, and parental history of HF. It is widely adopted in over 70 countries worldwide. Furthermore, it serves as the intervention threshold in many postmenopausal and glucocorticoid-induced osteoporosis treatment recommendations.^6,7

While other chronic inflammatory joint diseases, such as rheumatoid arthritis, are incorporated into FRAX calculations, no specific adjustment exists for axSpA. Research regarding the clinical utility of FRAX in axSpA patients remains limited. This study aims to examine the performance of FRAX with and without BMD using the proposed Thai cutoff values in predicting fragility fractures in axSpA patients and determine the appropriate cutoff of FRAX to improve the performance in axSpA patients.

Methods

Study design and participants

This cross-sectional study was conducted at Phramongkutklao Hospital, Bangkok, Thailand, from January 2020 to December 2023. Patients aged 40 years and older who fulfilled the Assessment of SpondyloArthritis International Society criteria (ASAS) for axSpA were consecutively recruited.⁸ Patients with secondary causes of osteoporosis, a history of cancer, chronic kidney disease (CKD) stage IV/V, hyperthyroidism, hyperparathyroidism, pregnancy, and/or lactation were excluded.

Demographic and clinical data were extracted, including age, sex, weight, height, body mass index (BMI), history of alcohol consumption, menopausal status, disease duration (years since diagnosis), and family history of axSpA. Medication history was also reviewed, including use of non-steroidal anti-inflammatory drugs, corticosteroids, proton pump inhibitors, conventional disease-modifying antirheumatic drugs (DMARDs), biologic DMARDs, and anti-osteoporotic drugs.

Laboratory data were collected, including complete blood count, erythrocyte sedimentation rate, C-reactive protein (CRP), serum calcium, phosphate, albumin, vitamin D levels, and HLA-B27 status.

Disease activity and function were assessed, including patient global assessment, the Thai version of the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI),⁹ the Thai version of Bath Ankylosing Spondylitis Functional Index (BASFI),⁹ the Bath Ankylosing Spondylitis Metrology (BASMI), and the Ankylosing Spondylitis Disease Activity Score (ASDAS).

The sample size was estimated based on the expected prevalence of osteoporotic fractures in patients with axSpA. Using an estimated prevalence of approximately 7.5% (based on Thai data²), a 95% confidence level, and a precision of 5%, the minimum required sample size was calculated to be 107 patients.

The study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.¹⁰

Fracture risk assessment tool

The 10-year probabilities of MOF (FRAX-MOF) and HF (FRAX-HF) were computed using the Thai database. A MOF was defined as a fracture of the hip, spine, wrist, or humerus.¹¹ Although the high-risk thresholds for postmenopausal osteoporosis are typically set at ⩾20% for FRAX-MOF and ⩾3% for FRAX-HF,¹² few patients in this cohort met the 20% threshold. Previous data from postmenopausal Thai women indicated that the optimal cutoff for FRAX-MOF without BMD was approximately 10%.¹³ Therefore, FRAX-MOF and FRAX-HF were categorized using cutoffs of 10% and 3%, respectively.

Bone mineral density

BMD and vertebral fracture assessment (VFA) were assessed at the time of enrollment. BMD was measured at the LS, FN, and total hip using the GE-Lunar iDXA and GE-Lunar DPX Duo densitometer (GE Healthcare, Madison, WI, USA). Osteoporosis was defined as a BMD T-score of −2.5 or less in postmenopausal women and men aged ⩾50 years or the presence of one or more fragility fractures. Whereas a Z-score of −2.0 or less was defined as being below the expected range for age (low BMD) in premenopausal patients and men aged <50 years.¹⁴ Fractured vertebrae were excluded from lumbar DXA analysis.

Radiography and VFA

The VFA by DXA and/or the lateral thoracolumbar X-rays were used to determine the presence of vertebral fracture, which was defined using Genant’s semiquantitative method and review by a certified nuclear radiologist.¹⁵ Sacroiliac joints were assessed by radiograph and graded for sacroiliitis according to the modified New York criteria.¹⁶

Statistical analyses

Categorical variables were presented as frequencies and percentages, while continuous independent variables were expressed as mean with standard deviation (SD) or median with interquartile range (IQR) for non-normally distributed variables. The differences between axSpA patients with and without osteoporotic fractures were assessed using the independent t-test for normally distributed continuous variables, Mann–Whitney U test for non-normally distributed continuous variables, Chi-Square test, and Fisher’s exact test for categorical variables, as appropriate. The p-values <0.05 were considered statistically significant. Missing data were addressed using multiple imputations by chained equations.

The predictive performance of FRAX-MOF/HF for osteoporotic fractures was evaluated using receiver operating characteristic (ROC) curves. The area under the ROC curve (AUC) was calculated to assess the overall discriminative ability. The Youden index (Sensitivity + Specificity − 1) was used to determine the optimal cutoff point for each version of FRAX-MOF and FRAX-HF. Sensitivity, specificity, false positive rate, Youden index, and odds ratios with 95% confidence intervals (CI) were calculated against actual MOF and HF.

Visual representations were used to enhance data interpretation. Bar charts were created to compare median FRAX scores between fracture and non-fracture groups. Scatterplots with regression lines were also constructed to explore the relationship between FRAX scores and age stratified by fracture status.

All statistical analyses were performed using Stata version 12 (StataCorp, College Station, TX, USA).

Results

Patient characteristics

A total of 158 consecutive axSpA patients were initially screened. Thirty-one patients were excluded: age under 40 years (n = 27), CKD stage 4 or 5 (n = 2), and a history of malignancy (n = 1). Consequently, 125 axSpA patients were enrolled in the final analysis (Figure 1). The majority were male (74%), with a mean (SD) age of 50.9 (10.8) years and a median (IQR) disease duration of 72.0 (20,191) months. HLA-B27 was detected in 85.7%. The BMD T-score indicated osteoporosis (⩽−2.5) was found in 13 (13.8%) patients. Osteoporotic fractures were identified in 20 (21.3%) patients, including 19 vertebral fractures, 1 HF, and 1 proximal humerus fracture. One patient experienced both vertebral and HF.

Figure 1.

Flow diagram of the study.

There were no statistically significant differences between axSpA patients with and without fracture regarding age, gender distribution, or median disease duration. Patients with fracture reported a higher mean (SD) BMI than those without fracture (26.1 (5.4) and 24.1 (4.5), respectively; p = 0.040), likely due to a lower mean (SD) height (160.7 (8.8) vs 164.5 (8.2); p = 0.031) and more median (IQR) height loss (2.5 (0, 8.0) vs 0 (0.0, 2.7); p = 0.014) in patients with fracture compared to those without fractures.

Patients with fractures had higher disease activity as assessed by BASDAI and ASDAS, poorer function as examined by BASFI, and less spinal mobility and flexibility as measured by BASMI than those without fractures. In addition, patients with fractures were more likely to receive anti-osteoporotic medications and had higher vitamin D levels due to vitamin D supplementation compared to those without fractures. There was no difference in axSpA treatment between the two groups. The baseline demographic and clinical characteristics, stratified by fracture status, are presented in Table 1.

Table 1.

Baseline characteristics according to fracture status.

Variables	All patients (n = 125)	No fracture (n = 95)	Fracture (n = 30)	p-Value
Demographic parameters
Age (years), mean (SD)	50.9 (10.8)	50.4 (10.7)	52.3 (11.3)	0.406
Male, n (%)	92 (73.6)	69 (72.6)	23 (76.7)	0.662
Weight (kg), mean (SD)	66.0 (14.8)	65.7 (15.0)	66.9 (14.3)	0.691
Height (cm), mean (SD)	163.6 (8.5)	164.5 (8.2)	160.7 (8.8)	0.031
Height loss, median (IQR)	1.0 (0.0, 4.0)	0.0 (0.0, 2.7)	2.5 (0.0, 8.0)	0.014
Body mass index(kg/m²), mean (SD)	24.6 (4.8)	24.1 (4.5)	26.1 (5.4)	0.040
Disease duration, median (IQR)	72.0 (19.5, 191.0)	63.0 (19.0, 180.0)	91.0 (22.0, 240.0)	0.282
History/current tobacco use, n (%)	41 (33.9)	33 (35.9)	8 (27.6)	0.411
History/current alcohol use, n (%)	43 (35.5)	35 (38.0)	8 (27.6)	0.305
HLA-B27 positivity, n (%)	104 (86.5)	80 (87.4)	24 (80.0)	0.467
Vitamin D, mean (SD)	30.0 (15.1)	26.2 (7.9)	38.9 (22.9)	0.007
Disease activity parameters
Patient global assessment, mean (SD)	25.0 (10.0, 50.0)	20.0 (10.0, 50.0)	40.0 (20.0, 70.0)	0.003
BASDAI, mean (SD)	2.7 (1.2, 4.3)	2.2 (1.0, 3.8)	4.2 (2.4, 5.0)	0.004
BASFI, mean (SD)	1.9 (0.8, 3.9)	1.6 (0.7, 3.7)	3.3 (1.8, 5.5)	0.011
BASMI, mean (SD)	3.6 (2.2, 6.0)	3.2 (2.0, 5.2)	6.7 (3.2, 7.6)	0.001
ASDAS CRP, mean (SD)	2.1 (1.2, 3.1)	1.7 (1.1, 2.7)	3.1 (2.0, 3.5)	<0.001
ASDAS ESR, mean (SD)	2.2 (1.6, 3.2)	2.1 (1.5, 3.0)	3.0 (1.9, 3.6)	0.009
Grade 4 sacroiliitis, n (%)	49 (39.8%)	33 (35.1%)	16 (55.2%)	0.054
Bone parameters
BMD at lumbar spine (g/cm²), mean (SD)	1.2 (0.2)	1.2 (0.2)	1.1 (0.2)	0.256
BMD at femoral neck (g/cm²), mean (SD)	0.8 (0.2)	0.9 (0.2)	0.8 (0.1)	0.316
L1–L4 T-score ⩽−2.5	6 (4.8%)	4 (4.2%)	2 (6.7%)	0.559
Femoral neck T-score ⩽−2.5	9 (7.5)	5 (5.4)	4 (14.8%)	0.237
Medication
Glucocorticoid, n (%)	8 (6.4%)	5 (5.3%)	3 (10.0%)	0.397
Anti-tumor necrosis factor, n (%)	46 (36.8%)	31 (32.6%)	15 (50.0%)	0.085
Anti-interleukin 17, n (%)	4 (3.2%)	3 (3.2%)	1 (3.3%)	1.000
Anti-osteoporotic drugs, n (%)	13 (10.4%)	5 (5.3%)	8 (26.7%)	0.003

ASDAS, Ankylosing Spondylitis Disease Activity Score; BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; BASFI, Bath Ankylosing Spondylitis Functional Index; BASMI, Bath Ankylosing Spondylitis Metrology Index; BMD, bone mineral density; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; SD, standard deviation.

FRAX performance and conventional postmenopausal treatment thresholds

The baseline 10-year fracture probabilities for MOF and HF, as estimated by FRAX according to fracture status, are summarized in Table 2 and illustrated in Figure 2. Patients who experienced fractures had significantly higher median FRAX-MOF scores than those without fractures, both when calculated without BMD (4.9 (IQR 2.7–7.0) vs 1.9 (IQR 1.3–3.6); p < 0.001) and with BMD (4.7 (IQR 3.2–7.9) vs 2.4 (IQR 1.6–4.8); p < 0.001). Similarly, median FRAX-HF scores were significantly higher in the fracture group, both without BMD (0.8 (IQR 0.3–2.1) vs 0.2 (IQR 0.1–0.9); p < 0.001) and with BMD (0.9 (IQR 0.4–2.2) vs 0.5 (IQR 0.1–1.2); p = 0.027).

Table 2.

Baseline FRAX 10-year risks according to fracture status.

Variables	All (n = 125)	No fracture (n = 95)	Fracture (n = 30)	p-Value
FRAX
FRAX-MOF without BMD, median (IQR)	2.3 (1.4, 4.7)	1.9 (1.3, 3.6)	4.9 (2.7, 7.0)	<0.001
FRAX-MOF with BMD, median (IQR)	3.0 (1.8, 5.4)	2.4 (1.6, 4.8)	4.7 (3.2, 7.9)	<0.001
FRAX-HF without BMD, median (IQR)	0.3 (0.1, 1.0)	0.2 (0.1, 0.9)	0.8 (0.3, 2.1)	<0.001
FRAX-HF with BMD, median (IQR)	0.6 (0.2, 1.6)	0.5 (0.1, 1.2)	0.9 (0.4, 2.2)	0.027
FRAX thresholds
FRAX-MOF without BMD ⩾10, n (%)	2 (1.6)	2 (2.1)	0 (0.0)	1.000
FRAX-MOF with BMD, ⩾10, n (%)	8 (6.7)	3 (3.2)	5 (18.5)	0.014
FRAX-HF without BMD ⩾3, n (%)	16 (12.8)	9 (9.5)	7 (23.3)	0.062
FRAX-HF with BMD ⩾3, n (%)	20 (16.7)	14 (15.1)	6 (22.2)	0.388

BMD, bone mineral density; FRAX, Fracture Risk Assessment Tool; HF, hip fractures; IQR, interquartile range; MOF, major osteoporotic fractures.

Figure 2.

Median FRAX scores in axSpA patients with and without fractures. (a) FRAX-MOF scores without and with BMD. The dashed line represents the standard clinical cutoff (⩾10%) for high fracture risk. (b) FRAX-HF scores without and with BMD. The dashed line represents the standard clinical cutoff (⩾3%) for high fracture risk.

However, most patients in both the fracture and non-fracture groups had FRAX probabilities below the conventional Thai postmenopausal treatment thresholds (⩾10% for MOF and ⩾3% for HF). When conventional postmenopausal FRAX treatment thresholds were applied, the ability to discriminate between patients with and without fragility fractures was substantially attenuated. Using the threshold for FRAX-MOF ⩾10%, only 1.6% of the overall cohort met the threshold when BMD was not included, and notably, none of the patients in the fracture group reached this cutoff (p = 1.000). When BMD results were incorporated into FRAX calculation, a higher proportion of patients with fractures fulfilled the MOF ⩾10% threshold, and more patients in the fracture group reached the threshold as compared with those without fractures (18.5% vs 3.2%, p = 0.014). However, the absolute number of patients identified remained small.

A similar pattern was observed for HF probability. Despite significantly higher median FRAX probabilities in patients with fractures, there were no statistically significant differences between fracture and non-fracture groups, either without BMD (23.3% vs 9.5%, p = 0.062) or with BMD (22.2% vs 15.1%, p = 0.388), suggesting that the conventional cutoff of FRAX-HF ⩾3% cannot distinguish patients with a fragility fracture from those without it. The data are presented in Table 2 and Figure 2.

Analysis of the scatterplots presented in Figure 3 demonstrated a progressive increase in FRAX probabilities with advancing age in both groups, and this trend appears steeper among patients who have had fractures. At comparable age categories, patients with fractures consistently had higher FRAX-MOF and FRAX-HF values than those without fractures. Notably, only a small subset of fracture patients aged over 60 exhibited FRAX-MOF scores approaching or surpassing the standard clinical cutoff of ⩾10%, and most patients with fractures had FRAX-HF below the standard clinical cutoff of ⩾3%, regardless of BMD inclusion.

Figure 3.

Relationship between age and FRAX scores according to fracture status. (a) FRAX-MOF scores without and with BMD. Dashed line represents the standard clinical cutoff (⩾10%) for high fracture risk. (b) FRAX-HF scores without and with BMD. Dashed line represents the standard clinical cutoff (⩾3%) for high fracture risk.

ROC curves for FRAX with and without BMD are presented in Figure 4(a). The AUC for FRAX-MOF without BMD was 0.770 (95% CI: 0.672–0.868), and FRAX-MOF with BMD demonstrated a slightly lower AUC of 0.725 (95% CI: 0.618–0.831). As shown in Figure 4(b), the AUCs for FRAX-HF without and with BMD were 0.674 (95% CI: 0.562–0.787) and 0.608 (95% CI: 0.496–0.721), respectively. These findings suggest that FRAX, particularly FRAX-MOF without BMD, has good discriminative performance for fractures; however, the conventional postmenopausal thresholds for both FRAX-MOF and FRAX-hip have limited sensitivity for identifying individuals with high risk for fractures and underestimate fracture risk in axSpA patients, particularly in younger individuals. Therefore, significantly lower thresholds are required for effective screening in this population.

Figure 4.

ROC curve of FRAX-MOF and FRAX-HF with and without BMD in relation to fragility fractures in axSpA patients. (a) FRAX-MOF without BMD: AUC = 0.770 (95% CI: 0.672–0.868); FRAX-MOF with BMD: AUC = 0.725 (95% CI: 0.618–0.831). (b) FRAX-HF without BMD: AUC = 0.674 (95% CI: 0.562–0.787; FRAX-HF with BMD: AUC = 0.608 (95% CI: 0.496–0.721).

Optimal FRAX threshold in AxSpA patients

The performance characteristics of the FRAX-MOF both with and without BMD are summarized in Table 3. Applying the standard high-risk cut-off for FRAX-MOF (⩾10%) had a sensitivity of 0.0% and a high specificity of 97.9%. With the inclusion of BMD, the ⩾10% threshold remained inadequate, providing a marginal sensitivity of 18.5% and a specificity of 96.8%. Notably, no subjects in this axSpA cohort reached the traditional treatment threshold of a FRAX-MOF ⩾20%, illustrating that these thresholds are not good for fracture screening and are clinically inapplicable for this population. In contrast, the optimal performance for FRAX-MOF without BMD was identified at a significantly lower cutoff of ⩾4%. This threshold yielded the highest Youden index (0.356), with a sensitivity of 56.7% and a specificity of 78.9%. When BMD was incorporated into the model, the ⩾4% cutoff remained the most effective, providing a Youden index of 0.281, a sensitivity of 59.3%, and a specificity of 68.8%.

Table 3.

The cut-off and performance characteristics of the 10-year-probability of MOF (FRAX-MOF with and without BMD) with sensitivity, specificity, false positive rate, and Youden index in relation to actual fragility fractures.

FRAX-MOF (10-year probability of MOF)	Without BMD				With BMD
FRAX-MOF (10-year probability of MOF)	Sensitivity	Specificity	False positive rate	Youden index	Sensitivity	Specificity	False positive rate	Youden index
⩾4	56.7	78.9	45.9	0.356	59.3	68.8	35.6	0.281
⩾5	50	85.3	51.7	0.353	44.4	75.3	34.3	0.197
⩾6	33.3	90.5	52.6	0.238	33.3	84.9	39.1	0.182
⩾7	26.7	91.6	50	0.183	29.6	90.3	47.1	0.199
⩾8	13.3	95.8	50	0.091	22.2	95.7	60	0.179
⩾9	10	95.8	42.9	0.058	18.5	96.8	62.5	0.153
⩾10	0	97.9	0	−0.021	18.5	96.8	62.5	0.153
⩾11	0	97.9	0	−0.021	18.5	97.8	71.4	0.163
⩾12	0	98.9	0	−0.011	14.8	97.8	66.7	0.126
⩾20	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A

BMD, bone mineral density; FRAX, Fracture Risk Assessment Tool; HF, hip fractures; MOF, major osteoporotic fractures; N/A, not applicable.

The performance characteristics of the FRAX-HF both with and without BMD are summarized in Table 4. The standard high-risk cutoff for FRAX-HF (⩾3%) had a sensitivity of 23.3% and a specificity of 90.5%. Similarly, the inclusion of BMD resulted in limited sensitivity at 22.2%, despite a good specificity of 84.9%. Optimal thresholds were identified at values significantly lower than traditional benchmarks. Without BMD, a threshold of ⩾1.5% achieved the maximal Youden index (0.253), corresponding to a sensitivity of 40% and a specificity of 85.3%, and a threshold of ⩾1% had a sensitivity of 46.7%, a specificity of 75.8%, and the Youden index of 0.225. With the inclusion of BMD, the highest Youden index (0.148) was observed at a threshold of ⩾1%, resulting in a sensitivity of 48.1% and a specificity of 66.7%.

Table 4.

The cutoff and performance characteristics of the 10-year-probability of HF (FRAX-Hip: with and without BMD) with sensitivity, specificity, false positive rate, and Youden index in relation to actual fragility fractures.

FRAX-Hip (10-year probability of HF)	Without BMD				With BMD
FRAX-Hip (10-year probability of HF)	Sensitivity	Specificity	False positive rate	Youden index	Sensitivity	Specificity	False positive rate	Youden index
⩾1	46.7	75.8	37.8	0.225	48.1	66.7	29.5	0.148
⩾1.5	40	85.3	46.2	0.253	29.6	75.3	25.8	0.049
⩾2	26.7	89.5	44.4	0.162	25.9	81.7	29.2	0.076
⩾2.5	23.3	89.5	41.2	0.128	22.2	83.9	28.6	0.061
⩾3	23.3	90.5	43.8	0.138	22.2	84.9	30	0.071
⩾3.5	23.3	92.6	50	0.159	22.2	88.2	35.3	0.104
⩾4	10	94.7	37.5	0.047	22.2	92.5	46.2	0.147
⩾4.5	3.3	94.7	16.7	−0.02	18.5	92.5	41.7	0.11
⩾5	3.3	96.8	25	0.001	14.8	93.5	40	0.083

BMD, bone mineral density; FRAX, Fracture Risk Assessment Tool; HF, hip fractures.

Discussion

The present study demonstrates that fragility fractures, especially vertebral fractures, are common in axSpA patients; however, they are often unrecognized and thus untreated. Only 30% of axSpA patients with fractures in this study received anti-osteoporotic drugs. This data emphasizes the importance of screening for osteoporosis and vertebral fractures in axSpA patients. FRAX is the most widely adopted and user-friendly tool for fracture risk assessment worldwide. FRAX scores in axSpA patients are generally lower than those of postmenopausal women due to their younger age and predominantly male gender. The mean FRAX-MOF and FRAX-HF scores of axSpA patients in this study (2.3%–3% and 0.3%–0.6%) were comparable to those reported in Korea (3.6% and 1.0%),¹⁷ Tunisia (0.36% and 0.3%),¹⁸ and India (1.53% and 0.54%)¹⁹; however, they are lower than those reported in Sweden (9.9% and 2.4%).²⁰ The variations can be attributed to the older age and higher prevalence of females, with approximately half experiencing menopause, among axSpA patients in the Swedish study, compared to those in other countries.^17–19

Our findings suggest that FRAX is an effective tool for discriminating fracture risk in patients with axSpA. This is supported by the significantly higher median FRAX scores observed in patients with fractures compared to those without fractures. Furthermore, the ROC analysis yielded a good discriminative performance, particularly for FRAX-MOF without BMD, which achieved an AUC of 0.770. These results indicate that the tool itself is capable of identifying high-risk individuals. However, the limitations are the inappropriately high conventional postmenopausal treatment thresholds. The standard FRAX thresholds for MOF (⩾20%) and HF (⩾3%)^12,19 were originally developed for postmenopausal women and the general population. Notably, no subject in this study reached the ⩾20% threshold for FRAX-MOF. Therefore, we categorized using the thresholds of 10% and 3% for FRAX-MOF and FRAX-HF, respectively, in accordance with earlier research in Thailand.^13,14 Despite lowering the threshold of FRAX-MOF to ⩾10%, the sensitivity is still very low in identifying patients with fragility fractures in this cohort.

Our findings suggest that FRAX underestimates the actual fracture risk in patients with axSpA. Likewise, a small study in 40 axSpA patients from India by Singh et al.¹⁹ found that although many axSpA patients had low BMD and high disease activity, their FRAX scores rarely exceeded intervention thresholds. The means (SD) of FRAX-MOF and FRAX-HF were only 3.5 (1.7) and 2.0 (1.3) in patients with high risk for fracture in axSpA patients as the BMD T-score osteoporosis by DXA, respectively.¹⁹ However, there was no report of the FRAX values in axSpA patients with fractures. These findings suggest that with the postmenopausal thresholds, the fracture risk would be profoundly underestimated in axSpA patients.

The discrepancy between the thresholds of FRAX scores between axSpA patients and postmenopausal women likely came from the unique pathophysiology of fractures in axSpA. While FRAX incorporates age and BMD, it does not capture risks associated with chronic systemic inflammation and mechanical structural changes in axSpA patients. This is supported by the data from this study and others that patients with fractures exhibited significantly higher levels of disease activity (ASDAS and BASDAI) and poorer functional indices (BASFI and BASMI) than those without fractures.^2,21 Chronic inflammation is a key driver of bone loss in axSpA. Inflammatory cytokines such as TNF-α and IL-17 can promote osteoclast activation and inhibit bone formation,²² thereby increasing fracture risk even in patients with normal BMD.

To improve the sensitivity of FRAX as a screening tool for fragility fractures in axSpA patients, we proposed the optimal thresholds for FRAX-MOF and FRAX-HF in axSpA patients as ⩾4% and HF ⩾1.5%, respectively. These lower thresholds provided a superior diagnostic balance between sensitivity and specificity (sensitivity 56.7%, specificity 78.9% for MOF; sensitivity 40%, specificity 85.3% for HF). The other approach by Mulkerrin et al.²³ was presented in an abstract, which reported that FRAX modification by imputing rheumatoid arthritis can enhance sensitivity and further identify axSpA patients who are at high risk of fractures. By adopting these disease-specific adjustments, clinicians can significantly improve the identification of high-risk individuals who would otherwise be missed by standard postmenopausal screening thresholds.

Among different FRAX calculations, FRAX-MOF without BMD demonstrated the best performance, achieving the highest Youden index and AUC (0.770). This is highly relevant as most fractures in axSpA are vertebral. Furthermore, the inclusion of BMD did not enhance the discriminative performance of FRAX. Syndesmophyte formation and ligamentous calcification can falsely elevate LS BMD readings. Hip osteoarthritis and avascular necrosis can further complicate hip BMD interpretation. Consequently, FRAX-MOF without BMD is more accurate in this context. Furthermore, it provides a practical, cost-effective screening method for fractures that do not require DXA.

Limitations

First, the sample size was relatively small and limited to a single center, potentially restricting generalizability. Second, FRAX probabilities were calculated using the Thai-specific model; the optimal thresholds in this study may vary across different geographic regions or ethnic groups with distinct baseline fracture risks. Third, LS BMD in axSpA may be falsely elevated due to syndesmophytes, potentially leading to underestimation of fracture risk. Lastly, while the FRAX-MOF without BMD performed the best, we did not incorporate advanced imaging like the TBS or a lateral view of LS BMD, which have been reported to improve the sensitivity of predicting vertebral fractures in axSpA patients.^2,24 The absence of these measures may have limited the predictive performance of our analysis.

Conclusion

FRAX is an acceptable tool for discriminating fracture risk in patients with axSpA. However, the conventional postmenopausal treatment thresholds demonstrate very low sensitivity. We recommend adopting axSpA disease-specific thresholds of ⩾4% for FRAX-MOF and ⩾1.5% for FRAX-HF without BMD. The FRAX-MOF without BMD showed the best performance, indicating that clinical risk factors alone may be a sufficient, practical, and cost-effective screening method for routine practice. Implementing these lower thresholds will enhance the identification of high-risk patients and facilitate timely bone-protective interventions in axSpA patients.

Supplemental Material

sj-docx-1-tab-10.1177_1759720X261447456 – Supplemental material for Underestimation of fracture risk by conventional FRAX thresholds in axial spondyloarthritis: a proposal for disease-specific thresholds

Supplemental material, sj-docx-1-tab-10.1177_1759720X261447456 for Underestimation of fracture risk by conventional FRAX thresholds in axial spondyloarthritis: a proposal for disease-specific thresholds by Sumapa Chaiamnuay and Apiwit Harirattanakul in Therapeutic Advances in Musculoskeletal Disease

Footnotes

Acknowledgements

The authors sincerely thank Professor Chatlert Pongchaiyakul for his insightful review and constructive comments which greatly contributed to the improvement of this manuscript and Ms. Sujittra Suriwong for performing statistical analyses.

Declarations

ORCID iDs

Sumapa Chaiamnuay

Apiwit Harirattanakul

Supplemental material

Supplemental material for this article is available online.

References

Molto

Nikiphorou

Comorbidities in spondyloarthritis. Front Med (Lausanne) 2018; 5: 62.

Saisirivechakun

Mahakkanukrauh

Pongchaiyakul

, et al. Prevalence of low trabecular bone score and its association with disease severity and activity in patients with axial spondyloarthritis. Sci Rep 2023; 13: 16258.

Pray

Feroz

Nigil Haroon

Bone mineral density and fracture risk in ankylosing spondylitis: a meta-analysis. Calcif Tissue Int 2017; 101: 182–192.

Prieto-Alhambra

Munoz-Ortego

De Vries

, et al. Ankylosing spondylitis confers substantially increased risk of clinical spine fractures: a nationwide case-control study. Osteoporos Int 2015; 26: 85–91.

Ward

Deodhar

Gensler

, et al. 2019 Update of the American College of Rheumatology/Spondylitis Association of America/Spondyloarthritis Research and Treatment Network Recommendations for the treatment of ankylosing spondylitis and nonradiographic axial spondyloarthritis. Arthritis Rheumatol 2019; 71: 1599–1613.

Charatcharoenwitthaya

Jaisamrarn

Songpatanasilp

, et al. Summary of the Thai Osteoporosis Foundation (TOPF) clinical practice guideline on the diagnosis and management of osteoporosis 2021. Osteoporosis Sarcopenia 2023; 9: 45–52.

Humphrey

Russell

Danila

, et al. 2022 American College of Rheumatology Guideline for the prevention and treatment of glucocorticoid-induced osteoporosis. Arthritis Rheumatol 2023; 75: 2088–2102.

Rudwaleit

Landewe

van der Heijde

, et al. The development of Assessment of SpondyloArthritis international Society classification criteria for axial spondyloarthritis (part I): classification of paper patients by expert opinion including uncertainty appraisal. Ann Rheum Dis 2009; 68: 770–776.

Kittiyanpanya

Chaiamnuay

Asavatanabodee

, et al. Reliability and validity of the Thai version of bath ankylosing spondylitis indices. J Med Assoc Thai 2014; 97: 381–385.

10.

Von Elm

Altman

Egger

, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. PLoS Med 2007; 4: e296.

11.

Briot

Paternotte

Kolta

, et al. FRAX(R): prediction of major osteoporotic fractures in women from the general population: the OPUS study. PLoS One 2013; 8: e83436.

12.

Eastell

Rosen

Black

, et al. Pharmacological management of osteoporosis in postmenopausal women: an Endocrine Society Clinical Practice Guideline. J Clin Endocrinol Metab 2019; 104: 1595–1622.

13.

Sribenjalak

Charoensri

Pongchaiyakul

An optimal intervention threshold of FRAX in postmenopausal Thai women. Arch Osteoporos 2022; 17: 21.

14.

Kanis

JA.

Assessment of fracture risk and its application to screening for postmenopausal osteoporosis: synopsis of a WHO report. WHO Study Group. Osteoporos Int 1994; 4: 368–381.

15.

Genant

van Kuijk

, et al. Vertebral fracture assessment using a semiquantitative technique. J Bone Miner Res 1993; 8: 1137–1148.

16.

Van der Linden

Valkenburg

Cats

. Evaluation of diagnostic criteria for ankylosing spondylitis. A proposal for modification of the New York criteria. Arthritis Rheum 1984; 27: 361–368.

17.

Kang

Kwok

, et al. Assessment of fracture risk in patients with axial spondyloarthritis: a case-control study using the fifth Korean National Health and Nutrition Examination Survey (KNHANES V). Scand J Rheumatol 2016; 45: 23–31.

18.

Henchiri

Hamdi

Maatallah

, et al. Assessment of fracture risk in patients with spondyloarthritis using the FRAX scores. Tunis Med 2019; 97: 1235–1239.

19.

Singh

Taneja

, et al. Assessment of fracture risk and its predictors in patients with axial spondyloarthritis using the fracture risk assessment (FRAX) algorithm. JIACM 2020; 21: 16–19.

20.

Klingberg

Lorentzon

Mellstrom

, et al. Osteoporosis in ankylosing spondylitis—prevalence, risk factors and methods of assessment. Arthritis Res Ther 2012; 14: R108.

21.

Liu

Dai

, et al. Prevalence and relative risk of fractures and osteoporosis in patients with ankylosing spondylitis: a systematic review and meta-analysis. Osteoporos Int. Epub ahead of print 23 January 2026. DOI: 10.1007/s00198-025-07838-x.

22.

Fassio

Atzeni

Rossini

, et al. Osteoimmunology of spondyloarthritis. Int J Mol Sci 2023; 24(19):14924.

23.

Mulkerrin

O’Shea

Fitzgerald

P164 Fracture risk in axial spondyloarthritis: a retrospective analysis of the use of the Fracture Risk Assessment Tool in a cohort of affected patients. Rheumatology 2024; 63: keae163.203.

24.

Lim

Kang

KY.

A contemporary view of the diagnosis of osteoporosis in patients with axial spondyloarthritis. Front Med (Lausanne) 2020; 7: 569449.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB