Ankylosing spondylitis (AS) is a chronic inflammatory disease of the axial skeleton. Cortical new bone formation leading to ankyloses of the spine and sacroiliac joints is a key characteristic of AS; however, AS is also associated with trabecular bone loss leading to osteoporosis [1,2]. Those opposite pathologic bone remodeling processes are an apparent paradox of this disease and take place in close proximity within the spine [1]. Indeed, patients with AS present with osteoporosis more frequently than the general population [2,3]. In addition, the characteristic osteoproliferation that increases the rigidity of the spine and bone loss in the vertebrae contributes to increased vulnerability of the spine in patients with AS. Previous studies have demonstrated that patients with AS have an increased risk of vertebral fracture (VF) that can be complicated by neurologic injuries [4,5]. Therefore, it is crucial to identify and selectively treat AS patients at risk for fractures.

Bone strength is determined by bone geometry, cortical thickness, porosity, trabecular bone morphology, and the intrinsic properties of bony tissue [6]. In clinical practice, bone strength is usually estimated indirectly from bone mineral density (BMD) determined using dual-energy X-ray absorptiometry (DXA) and the World Health Organization (WHO) definitions of osteopenia and osteoporosis [7]. Previous studies have found low BMD to be a good predictor of future fracture risk [8,9]. However, BMD corresponds to the ratio between the bone mineral content and the scanned bone area; it does not reflect bone architecture or quality. Moreover, it can be difficult to interpret lumbar spine BMD measured by DXA in the anteroposterior (AP) projection in advanced AS patients due to the presence of characteristic syndesmophytes [2,10-12]. Syndesmophytes develop in parallel to the progression of AS and thus mask the bone loss in AP spinal measurements of BMD [13]. Therefore, the European League Against Rheumatism recommended that patients with axial spondyloarthropathy (axSpA) who present with syndesmophytes on conventional lumbar spine radiography be assessed for osteoporosis by a hip DXA scan, supplemented by spine DXA in the lateral projection or possibly quantitative computed tomography (QCT) of the spine [14,15]. The trabecular bone score (TBS) is a recently developed tool for bone strength assessment that represents the textural parameter of pixel gray-level variations in lumbar spine DXA images [16]. Previous studies have shown that TBS could predict fractures independently of BMD and has additive value for discriminating patients with VF risk when combined with BMD [17-20]. In studies conducted in male AS patients [21] and patients with axSpA [11,12], TBS was not influenced by the characteristic syndesmophyte formation on the spine, unlike AP spine DXA. However, the data regarding the use of TBS in patients with AS was limited to a single study [21].

We aimed to reveal the usefulness of TBS for assessing bone strength in patients with AS in comparison with DXA methods (AP and lateral spine DXA and hip DXA) in a single-center AS cohort.

Among the 234 patients recruited for this study, 215 were included; eight patients did not complete the required physical exams, questionnaires, or BMD exam. One patient was eventually diagnosed with diffuse idiopathic skeletal hyperostosis rather than AS. Ten patients with a history of lumbar spine vertebroplasty or both hip replacement surgery were excluded from the study.

Our study population reflects the characteristic clinical differences between male and female patients with AS in Korea. In addition, TBS showed comparable predictability for detecting FRAX-MOF score of moderate or higher fracture risk compared to hip BMD or lumbar lateral BMD in the analyses of ROC curves in AS patients aged 40 years or older. This indicates that a low TBS could be associated with an increased risk of MOF that includes VF in patients with AS. The most distinctive finding of this study is that only TBS negatively correlated with mSASSS in both male and female AS patients. In other words, TBS was the only bone strength measurement that was unaffected by the characteristic spinal osteoproliferation of AS, even in male patients with advanced spinal changes. These key findings attest to the usefulness of TBS in assessing bone strength in patients with AS.

Many previous studies have shown that clinical characteristics could differ between male and female patients. Male AS patients generally show more prevalent HLA B27 positivity and advanced spinal radiographic change. In contrast, female patients present with more prevalent peripheral joint involvement and higher symptomatic burden of AS [29]. Our study showed similar characteristic differences between male and female AS patients, even though age and disease duration did not differ between the groups (Table 1). However, the similar BASDAI and C-reactive protein (CRP) found in female patients in this study differed from the findings of previous studies, which mainly included white ethnicity populations. BASDAI and CRP were higher in female patients in those previous studies [29]. As shown in Fig. 1, the distribution of patients according to mSASSS is distinctively right skewed, and at all spine levels, the mean grades of mSASSS were higher in male patients than in female patients. Thus, sex seems to play a very important role in AS and should always be regarded as an important confounding factor when investigating patients with AS. However, no previous study has separately evaluated correlations between bone strength measurement methods and the degree of spinal radiographic change in male and female AS patients [21].

The prevalence of VF history in this study was 4.2%, which was lower than the prevalence of 13.6% in the general population of a rural Korean community [30]. In a Swedish cohort of 204 AS patients by Klingberg et al. [15], only three patients (1%) reported previous VF. However, 24 patients (11.8%) were found to have VF after the systemic radiographic evaluation. The presence of VF was also determined by systemic radiographic evaluation in the cohort of a Korean rural community [30]. The ankylosed spine of AS patients is prone to fracture even after minor trauma, and spinal fractures are asymptomatic in some patients, leading to delay in diagnosis. Furthermore, X-ray of the spine in patients with AS may be difficult to interpret owing to the preexisting spinal structural change of AS [5]. Klingberg et al. [15] also showed that the prevalence of VF varied among studies from 9% to 42% depending on different definitions of VF, making a straightforward comparison between studies impossible. The prevalence of non-VF in our study was higher than in the Swedish cohort result (21.9% vs. 9%) [15]. However, both studies have limitation for comparison of fracture prevalence as traumatic causes were not distinguished.

In general, males tend to show higher BMD compared to females both at the lumbar spine and hip [31]. Measurement of BMD in this study was also higher in male patients at every examined site (Table 2). But, TBS is a pure bone texture index that measures different aspects of bone strength independent of BMD [16,32]. No significant difference in TBS between the sexes was noted in this study, like the results of a previous study conducted in the general population [32]. Indeed, there was no previous study to compare with our results regarding the TBS differences between male and female AS patients. However, patients with low TBS (≤ 1.31) were more prevalent in female in patients aged 50 years or older, as shown in Table 2. Female AS patients aged 50 years or older could be at higher risk of fracture compared to male patients of same age group in aspects of bone density and trabecular bone quality. However, predicting future fracture risk is a complicated task because many clinical factors other than BMD can contribute to it. Those clinical factors include age, sex, prior fragility fractures, parental history of hip fracture, smoking history, alcohol consumption, long-term glucocorticoid use, and certain comorbidities such as rheumatoid arthritis [33]. FRAX is an algorithm developed by WHO that calculates 10-year fracture probability using those clinical factors. Our results show that hip BMD, lumbar lateral BMD, and TBS predict a moderate or higher risk of FRAX-MOF comparably well in patients aged 40 years or older. FRAX-MOF reflects the probability of fracture in the hip, clinical spine, wrist, or humerus [33]. TBS was not inferior in predicting MOF compared to hip DXA scan in this group of AS patients. In a previous study of patients with axSpA, TBS was lower in patients with VF than in those without VF, and TBS showed better discriminatory value than total hip BMD in the prediction of VF [34]. Our result, taken together with previous reports, support the idea that TBS could predict future VF in patients with AS. However, FRAX has the limitation of not considering many causes of secondary osteoporosis [33]. According to a study on the clinical risk factors for VF in AS patients, the presence of VF was significantly associated with higher mSASSS [35]. This result suggests that the risk of VF not only increases with the presence of the disease itself, but also with more advanced spinal structural change of AS. Therefore, we need to improve the clinical tools available to predict the future risk of VF in AS patients. The presence of AS and the severity of its spinal change could be good candidate clinical risk factors to consider when predicting the risk of VF in patients with AS [4,5].

Our study showed some contradictory results compared with two previous studies regarding the correlations between bone strength measurement methods and mSASSS. Femoral neck, total hip, and lumbar lateral BMD negatively correlated with mSASSS in the Swedish cohort of 204 AS patients [2]. In a Korean cohort of 100 male AS patients aged 50 years or younger by Kang et al. [21], femoral neck and total hip BMD also negatively correlated with SASSS, another radiographic scoring system that measures only lumbar spinal changes in AS patients. However, in our study, lumbar lateral BMD and hip BMD correlated negatively with mSASSS only in female patients, not in male patients. The Swedish group reported that their five-year follow-up data showed unexpected increase in lumbar lateral and total hip BMD but decrease in femoral neck BMD when mSASSS increased. Because the increase in lumbar AP and lateral BMD was significant only for male patients, who had higher mSASSS than female patients overall, they suspected that the lumbar lateral DXA scan might also have been influenced by osteoproliferation of spine [36]. Male patients in our study included patients with much higher mSASSS than the patients in the Swedish cohort (mSASSS of 32.5 ± 23.1 in this cohort vs. 16.6 ± 20.9 in the Swedish cohort) and the patients in the Korean cohort conducted by Kang et al. [21] (SASSS of 13.71 ± 13.26 in this cohort vs. 9.5 ± 16.0 in the young male Korean AS cohort). Therefore, we also assume that advanced spinal structural changes could have affected some part of the lumbar spine and increased the lumbar lateral BMD. The increase in total hip BMD after 5 years of follow up in the Swedish cohort was also significant only in male patients (p < 0.001). Zhao et al. [37] investigated the factors associated with severe radiographic hip involvement in an AS cohort and identified spinal involvement as an associated clinical factor. Accompanying hip cortical changes, such as osteophyte formation, could have increased both total hip and femur neck BMD in male patients with advanced spinal disease in our cohort. Future studies that measure radiographic hip change or perform a QCT evaluation of both hip and lumbar spine could help to clarify these correlations.

One of the most notable findings in this study is that only TBS correlated negatively with mSASSS in both male and female AS patients. Though TBS also correlated negatively with SASSS in the cohort of young male AS patients by Kang et al. [21], our finding signifies that TBS is not influenced by the spinal osteoproliferation characteristic of AS, even in male patients with advanced spinal changes and female AS patients. The strengths of TBS include easy accessibility and low cost because it is calculated using lumbar AP DXA scan images [16]. In addition, compared with lumbar lateral DXA scans or QCT, TBS has fewer problems with poor precision or radiation hazards, respectively [2,38]. Because only a single study previously used TBS for patients with AS [21], more studies on this topic are needed, and TBS should be considered as an additional recommended tool for evaluating osteoporosis in patients with AS.

Our study has several limitations. First, we investigated the prevalence of fracture primarily by relying on patient history, without systemic X-ray evaluations for VF. However, we reviewed all the medical records of spinal X-ray interpretation to minimize this weakness. In addition, three patients with history of lumbar surgery were excluded in this study to facilitate accurate bone strength evaluation. Therefore, our result could underestimate the true prevalence of VF to some extent. Second, we did not examine the menopausal status in female patients, which has a significant impact on female bone health [39]. Park et al. [40] reported the mean ± SD age of natural menopause in Korea as 49.3 ± 0.07 years through the evaluation of 12,761 females using the Korea National Health and Nutrition Examination Survey data. Therefore, in the evaluation of patients with abnormal bone strength (Table 2), we limited female patients to aged 50 years or older as most of them are likely to be in postmenopausal or late perimenopausal period, when bone loss is accelerated owing to the hormonal change [39]. Third, because we did not perform QCT evaluations, the exact degree of cortical or trabecular change of the spine or hip was not evaluated. Lastly, mSASSS was measured by a single investigator at two different times. However, the intraobserver reliability between the two measurements was excellent (intraclass correlation coefficients, 1.00; p < 0.001).

In summary, among the bone strength evaluation methods examined in this study, only TBS was not influenced by the characteristic spinal osteoproliferation of AS in both male and female patients. In addition, TBS was not inferior in predicting the risk of MOF compared to hip BMD or lumbar spine lateral BMD. Lumbar spine AP BMD was inferior in predicting the risk of MOF compared to other methods. Our results indicate that TBS is a useful tool in assessing bone strength in patients with AS.