VERTEBRAL BONE MINERAL DENSITY, VERTEBRAL STRENGTH, AND SYNDESMOPHYTE GROWTH IN ANKYLOSING SPONDYLITIS: THE IMPORTANCE OF BRIDGING

Abstract

Objective:

To examine the relationship between vertebral trabecular bone mineral density (tBMD), vertebral strength, and syndesmophytes in ankylosing spondylitis (AS) using quantitative computed tomography (QCT).

Methods:

We performed spine QCT to measure syndesmophytes and tBMD in five vertebrae (T11 to L3) in 61 patients with AS. We used finite element analysis to measure vertebral strength in compressive overload, including in trabecular and cortical compartments. In cross-sectional analyses, we examined associations of syndesmophyte height with tBMD and vertebral strength in each vertebra. In 33 patients followed for two years, we examined if baseline tBMD and vertebral strength predicted syndesmophyte growth in the same vertebra, and vice versa.

Results:

In the cross-sectional analysis, 126 vertebrae had bridging, 77 vertebrae had non-bridging syndesmophytes, and 83 vertebrae had no syndesmophytes. There were strong inverse associations between syndesmophyte height and tBMD, total strength, and trabecular strength only among bridged vertebrae. In the longitudinal analysis, non-bridged vertebrae with low tBMD (adjusted beta −0.01; 95% confidence interval (CI) −0.019, −0.0012) and low strength (adjusted beta −0.0003; 95% CI −0.0004, −0.0002) had more syndesmophyte growth over time. Similar associations were absent among bridged vertebrae. Conversely, bridged vertebrae at baseline had a significant loss in percent tBMD over time (adjusted beta −0.001; 95% CI −0.0017, −0.0004).

Conclusion:

Associations between syndesmophytes and vertebral density and strength in AS differ between bridged and non-bridged vertebrae. Among non-bridged vertebrae, low tBMD and strength are associated with syndesmophyte growth. Bridging is associated with large subsequent losses in tBMD, possibly due to mechanical off-loading.

Syndesmophyte formation is a hallmark of ankylosing spondylitis (AS), but AS is also often accompanied by vertebral osteopenia.[1] The opposing pathological processes of bone formation and bone loss can occur simultaneously and in close proximity in the spine, in a manner possibly unique to AS. Both processes may lead to structural damage and functional consequences. Syndesmophytes may lead to bridging and loss of flexibility, whereas trabecular bone loss can predispose to vertebral fractures.[2,3] Both processes also share the risk factors of smoking and inflammation, and although data on the bone effects of nonsteroidal anti-inflammatory drug (NSAID) treatment in AS are limited, treatment with tumor necrosis factor inhibitors (TNFi) can improve bone mineral density (BMD) and may slow syndesmophyte development.[4–9]

Despite these commonalities, whether, and how, these processes are related is not well understood. It has been hypothesized that bridging syndesmophytes may promote bone loss by reducing vertebral mobility.[10] However, two longitudinal studies have suggested the converse, that low bone mineral density (BMD) predicts syndesmophyte formation.[11,12] Bone loss has also been proposed to result in vertebral instability, which is then hypothesized to promote syndesmophyte development as a compensatory mechanism.[13] Understanding this association is important because it would indicate whether treatments aimed at maintaining or improving BMD may also affect the progression of spine fusion.

Investigations of the relationship between syndesmophyte growth and bone loss in AS have been limited by the methods used to measure both syndesmophytes and BMD. Dual energy x-ray absorptiometry (DXA), a two-dimensional modality, is not valid in many patients with AS, because the anteroposterior projection can include syndesmophytes and posterior vertebral elements, often leading to the counterintuitive result that vertebral BMD increases with time.[14] Studies using lateral DXA have also reported increases in BMD over time.[15,16] In contrast, quantitative computed tomography (QCT), a three-dimensional modality, can selectively measure trabecular BMD (tBMD), and its measurements correlate with bone histomorphometry.[17] Similarly, the assessment of syndesmophytes has relied on plain radiography, which like DXA is a two-dimensional projection with problems of superimposition. Additionally, the accepted reading method, the modified Stoke AS Spinal Score, only examines syndesmophytes at anterior vertebral corners.[18] QCT is an alternative modality with better accuracy in syndesmophyte detection along the entire vertebral rim and better sensitivity to change.[19,20]

To date, no study has used CT to measure both vertebral BMD and syndesmophytes. Previous studies that measured BMD using QCT relied on radiography for detecting syndesmophytes. [10,17,21–26] Most studies used DXA for BMD and radiography for syndesmophytes. [11,12,15,16,27–29] Most studies also examined the lumbar spine globally, although an analysis of associations within individual vertebrae may provide more mechanistic insights. In addition, study of the mechanical strength of vertebrae may provide additional understanding of syndesmophyte development, beyond that provided by BMD. Vertebral strength is a measure of the breaking force of the vertebra under the action of a compressive overload, and is determined not only by BMD but also three-dimensional geometry, local variations of cortical and trabecular bone, and the spatial distribution of bone density.[30] Mechanistically, vertebral strength may be more closely related to syndesmophyte development than BMD. It is possible, for example, that the bone strength can be maintained in the face of reduced trabecular BMD if there are geometric or other morphological adaptations that provide alternative mechanisms of strength to the overall vertebral body. No prior studies have used finite element analysis to examine the relationship of vertebral strength with syndesmophytes in AS.

Our goal was to investigate the relationship between vertebral BMD, strength, and syndesmophytes using QCT.[31] In cross-sectional analyses, we examined the specific associations of volumetric tBMD, and trabecular and cortical bone strength, with syndesmophyte involvement in the same vertebra. In longitudinal analyses, we examined if baseline tBMD and vertebral strength predicted syndesmophyte growth, or if baseline syndesmophyte involvement predicted subsequent loss in vertebral tBMD. We hypothesized that bridging has a mechanical off-loading effect on vertebral trabecular bone, and therefore examined if associations differed between bridged and non-bridged vertebrae.

METHODS

Patients

Patients were enrolled in a prospective study of the use of spine CT to quantitate syndesmophytes in AS.[19,20] Inclusion criteria were: age ≥ 18 years, presence of AS by the modified New York criteria, and absence of extensive lumbar spine fusion by radiography (Bath AS Radiology Index (BASRI) < 4).[32] The study protocol was approved by the National Institute of Arthritis and Musculoskeletal and Skin Diseases institutional review board (#04-AR-0205), and all patients provided written informed consent.

The main study was a prospective longitudinal study with spine CT scans at baseline, one year, and two years, and clinical assessments every four months. We enrolled at least five subjects in BASRI categories 0, 1, 2, and 3 to include patients with a range of syndesmophyte involvement.[32] Other patients were enrolled in substudies that required a single visit with a spine CT scan. In all cases, clinical assessments included physical examination, medication histories, and collection of patient-reported measures. We computed the alternative AS Disease Activity Score (ASDAS) based on the Bath AS Disease Activity Index (BASDAI) and C-reactive protein levels.[33]

CT scanning

Patients enrolled early in the study (62%) were scanned on a Philips Brilliance 64 (slice thickness 1.5 mm) or GE Lightspeed Ultra (slice thickness 1.25 mm) scanner. More recent patients (38%) were scanned using a Siemens Somatom Flash or Somatom Force (slice thickness 1.0 mm) scanner. Estimated average absorbed radiation dose was 8.01 mSv.

We computed syndesmophyte height around the vertebral rim using a validated semi-automated method of high reliability and sensitivity to change.[31] In each disk space, the 360 degrees of the vertebral rim was divided into 72 angular sectors of 5 degrees each. In each angular sector, height was computed for ascending and descending syndesmophytes and normalized to the local IDS height so that bridging had a value of 1, values between 0 and 1 represented the proportion of the disk space spanned by syndesmophytes, and 0 indicated the absence of syndesmophytes. In theis study, we were interested in all syndesmophytes originating from a vertebral body rather than in each disk space. Therefore, for each vertebral body, we added all ascending syndesmophytes from the upper endplate, descending syndesmophytes from the lower endplate, and all bridging syndesmophytes from either endplate. The sum is a score termed syndesmophyte height, which has a possible range of 0 to 144, with 144 indicating total fusion with both neighboring vertebrae. Extent of bridging was the number of angular sectors with a score of 1 on either the upper or lower endplate.

Initial scans were performed from T10 to L4, while later scans also included the thoracic spine. To standardize the levels examined, we limited the analysis to five vertebrae (T11, T12, L1, L2, and L3) available for all patients. Data from T10 and L4 were needed to compute syndesmophyte heights for T11 and L3, respectively.

We used disc height loss, as seen on the lateral view, by the semiquantitative Videman scale to identify degenerative disc disease.[34] We classified discs with a score of 2 (disc narrower than the adjacent superior disc) or 3 (endplates almost touching) as having degeneration.

Volumetric trabecular bone mineral density

tBMD was calculated for each vertebral body (in mg/cm³) by relating measured voxel intensity to a calibration phantom scanned with each patient. The calculation was performed using commercial software (QCT-Pro, v 6.1, Mindways Software Inc., Austin, TX). This FDA-cleared software allows the user to select a region of interest of trabecular bone for which tBMD is computed (Supplemental figure 1). tBMD measurements were done by investigators blinded to syndesmophyte results. Intra-rater reliability for tBMD measurements, tested on 20 randomly selected vertebrae, was 0.974.

Finite element analysis by Biomechanical CT (BCT)

Finite element analysis was done on scans of 25 patients in the longitudinal study, selected to include a range of syndesmophyte involvement. BCT performs a finite element analysis to compute a measurement of bone strength, which is the force (in units of newtons) required to virtually break the patient’s bone in a standardized loading configuration. This type of virtual stress test combines image processing, bone biomechanics, and the well-established engineering structural analysis technique of non-linear finite element analysis to simulate a typical fracturing event: a compressive overload on the vertebra (Supplemental figure 2). BCT testing was performed with the VirtuOst software (version 1.2) by O.N. Diagnostics (Berkeley, CA, USA), which is FDA-cleared both to assess fracture risk and monitor bone loss and treatment. Details of these assessments can be found elsewhere.[30] Briefly, the target vertebra was segmented from the CT image, registered to a standardized orientation for loading, and the voxels were calibrated to standardized density units. To construct the finite element model, the bone was converted to 1mm cube-shaped elements, and each element was assigned element-specific elastic and failure properties. Vertebral total strength was then defined as the simulated force when applying a uniform compression to the vertebral body to a strain of 2%.

A separate finite element analysis was then performed to compute trabecular strength of each vertebral body. This was done by first virtually removing the cortex and subjecting the remaining bone to the same compressive overload. The contribution from the cortical compartment was then computed as the difference between whole-bone total strength and trabecular strength.

Statistical analysis

We performed two main analyses: a cross-sectional analysis using baseline data of all patients, and an analysis of changes over time using data of patients in the longitudinal protocol. In both cases, the unit of analysis was the individual vertebra.

In the cross-sectional analysis, we first examined the association between tBMD (the independent variable) and syndesmophyte height (the dependent variable) using regression analysis. This analysis tested whether vertebrae with lower tBMD were more likely to have extensive syndesmophytes than vertebrae with higher tBMD. We implemented the regression models as generalized estimating equations to account for the non-independence of vertebrae from the same patient, and adjusted for patient age, sex, smoking status (former/current versus never), vertebral level, current NSAID use, current TNFi use and C-reactive protein (CRP) level. We performed stratified analyses to investigate if the association between tBMD and syndesmophytes was modified by the presence of bridging. Models with alternative functional forms of tBMD (e.g. linear or quadratic terms) were compared for goodness of fit using the quasi-likelihood under the independence model criterion.

We repeated these analyses using data of the finite element analysis. We examined vertebral total strength, trabecular strength, cortical strength, and proportion of cortical strength contribution to total strength (cortical strength/total strength) as separate independent variables and syndesmophyte height as the dependent variable. We used generalized estimating equation models adjusted for patient age, sex, smoking, vertebral level, current NSAID use, current TNFi use, and CRP, and stratified the analyses by whether any bridging was present.

In the longitudinal analyses, we first examined the association of both tBMD and vertebral strength at baseline with change in syndesmophyte height in the same vertebra over the subsequent 2 years using a multiple regression model. Next, we examined the association of syndesmophyte height at baseline with percent change in tBMD in the same vertebra over the subsequent 2 years. Comparison of these results would indicate if tBMD was more likely to predict or follow syndesmophyte growth. These models were again implemented as generalized estimating equations, with adjustment for patient age, sex, smoking, vertebral level, NSAID use, TNFi use, and mean ASDAS on 7 visits over the two years. We did not include data from the CT scans at one year. We adjusted for ASDAS rather than CRP in the longitudinal analyses to incorporate aspects of patient-reported assessment. We repeated these analyses after excluding vertebrae with adjacent degenerated discs.

We considered associations to be statistically significant if the 95% confidence interval (CI) of regression coefficients excluded 0. We used SAS programs (version 9.4) for analysis.

RESULTS

Study cohort

We studied 61 patients in the cross-sectional analysis, 25 of whom were included in the finite element analysis. Thirty-three patients participated in the longitudinal study. Most patients were young adult or middle-aged men, and the mean duration of AS was 18.2 years (Table 1). Mean disease activity was low by BASDAI and moderate by ASDAS. Twenty-one patients (34.4%) had been treated with TNFi for a median (25^th, 75^th percentile) of 25 (13, 62) months. In the longitudinal cohort, 6 patients (18.2%) were on treatment with TNFi. No patients had been treated with bisphosphonates, other antiresorptive or anabolic medications, or biologics other than TNFi.

Table 1.

Patient characteristics at study entry. Plus-minus values are mean ± standard deviation. All others are n (%), except as noted.^*

Feature	All patients (N = 61)	Finite element analysis (N = 25)	Longitudinal sample (N = 33)
Age, years	45.1 ± 11.3	46.3 ± 11.2	45.8 ± 11.9
Men	52 (85.3)	23 (92.0)	28 (84.8)
White	52 (85.3)	21 (84.0)	28 (84.8)
Black	3 (4.9)	1 (4.0)	2 (6.1)
Asian	6 (9.8)	3 (12.0)	3 (9.1)
Non-smoker	40 (65.6)	17 (68.0)	21 (63.7)
Former smoker	19 (31.1)	7 (28.0)	11 (33.3)
Current smoker	2 (3.3)	1 (4.0)	1 (3.0)
Duration of AS, years	18.2 ± 11.2	19.9 ± 11.1	20.6 ± 12.3
BASDAI (0 – 10)	2.9 ± 1.9	2.9 ± 2.1	2.9 ± 1.9
BASFI (0 – 100)	26.2 ± 22.7	26.4 ± 23.1	25.3 ± 21.2
C-reactive protein, mg/L median (25th, 75th)	3.0 (1.7, 4.4)	0.4 (0.4, 11.2)	5.4 (1.4, 9.0)
AS Disease Activity Score	1.9 ± 0.8	1.9 ± 1.0	2.1 ± 0.8
Current nonsteroidal anti-inflammatory use	44 (72.1)	17 (68.0)	25 (75.8)
Current sulfasalazine use	3 (4.9)	1 (4.0)	2 (6.1)
Current tumor necrosis factor inhibitor use	22 (36.1)	5 (20.0)	6 (18.2)
Current prednisone use	3 (4.9)	1 (4.0)	1 (3.0)
Lumbar mSASSS, median (25th, 75th) (0 – 36)	4 (0, 8.5)	4 (0, 10)	4 (0, 9)
Syndesmophyte height, all vertebral levels (0 – 144)	17.2 ± 28.1	23.1 ± 35.0	18.4 ± 31.4
Any bridging	45 (73.7)	21 (84.0)	23 (69.7)
Trabecular bone mineral density, all vertebral levels, mg/cm3	131.9 ± 31.8	126.7 ± 31.2	132.5 ± 32.1
Total strength, all vertebral levels, N		8611 ± 2205
Trabecular strength, all vertebral levels, N	-	4753 ± 1423	-
Cortical strength, all vertebral levels, N	-	3812 ± 867	-
Proportion cortical strength (%), all vertebral levels	-	45.0 ± 5.4	-

^*AS = Ankylosing Spondylitis; BASDAI = Bath Ankylosing Spondylitis Disease Activity Index; BASFI = Bath Ankylosing Spondylitis Functional Index; mSASSS = modified Stoke Ankylosing Spondylitis Spine Score.

Cross-sectional associations with tBMD

Data on syndesmophyte height were available for 286 of 305 vertebrae in the cross-sectional analysis. Syndesmophyte height could not be calculated for 12 vertebrae (in 6 patients) because of extensive disciitis, and 7 patients were missing data on T11 because T10 was not completely in the field of view. Mean syndesmophyte height among all vertebrae was 17.2 ± 28.1 units (Table 1). Eighty-three vertebrae (29%) had no syndesmophytes, 77 vertebrae (26.9%) had only non-bridged syndesmophytes, and 126 vertebrae (44%) had a bridging syndesmophyte in at least one sector. Syndesmophyte heights and the frequency of bridging were greater at T11 and T12 than in the lumbar vertebrae. Mean tBMD was 131.9 mg/cm³ among all vertebrae.

There was an inverse association between tBMD and syndesmophyte height at each vertebral level, with Spearman correlations of −0.59, −0.53, −0.45, −0.32, and −0.49 at T11 to L3, respectively. The inverse association was also present in pooled data (Figure 1), including after adjustment for patient age, sex, smoking, vertebral level, NSAID use, TNFi use, and CRP (Supplemental table 1). The association was best fit as a curvilinear relationship, with little association between tBMD and syndesmophyte height at high tBMD, but a much stronger association at tBMD less than 120 mg/cm³. However, several vertebrae with low tBMD had no or only small syndesmophytes. TNFi use was not associated with syndesmophyte height. Results were similar if adjusted for duration of AS instead of age (Supplemental table 1).

Figure 1.

Association between trabecular bone mineral density (tBMD) and syndesmophyte height at baseline in 286 vertebrae from 61 patients. The top left panel shows all vertebrae, the top right panel shows non-bridged vertebrae, the bottom left panel shows bridged vertebrae, and the bottom right panel shows the association with the number of bridged sectors (possible range 1 to 144) among bridged vertebrae (n = 126).

When stratified by the presence of bridging, there was no significant association between tBMD and syndesmophyte height among vertebrae that were not bridged (n = 160) (Figure 1 and Supplemental table 1), although an inverse association was evident when adjusted for duration of AS instead of age. In contrast, there was a strong inverse curvilinear association among vertebrae with bridging (n = 126). Mean duration of AS was somewhat higher among patients contributing to the group with bridging (19.4 years versus 16.7 years), as was mean age (48.4 years versus 42.7 years) and the proportion of men (90.5% versus 81.2%).

There was a similar inverse curvilinear association between tBMD and the extent of bridging among vertebrae with any bridging (Figure 1 and Supplemental table 1). There was little association among vertebrae with tBMD > 120 mg/cm³, while vertebrae with tBMD < 120 mg/cm³ were more likely to have more extensive bridging. Among bridged vertebrae, there was no association between tBMD and the number of vertebral levels that were bridged, suggesting that tBMD was not influenced by bridging in neighboring vertebrae (Supplemental table 2).

Cross-sectional associations with Vertebral Strength

Data were available for 119 of 125 vertebrae (58 without bridging and 61 with bridging). Among all vertebrae, total strength was 8611 ± 2205 N (Table 1). Vertebral strength was directly correlated with tBMD, with a slightly higher correlation among bridged than non-bridged vertebrae (Supplemental figure 3).

In unadjusted analyses, syndesmophyte height was inversely associated total vertebral strength (Spearman r = −0.33), trabecular strength (r = −0.41), and cortical strength (r = −0.19) (Supplemental table 3). These associations were stronger among bridged than non-bridged vertebrae. Syndesmophyte height was directly correlated with proportion cortical strength among all vertebrae (r = 0.37), both in bridged and non-bridged subsets.

In the pooled adjusted analyses of non-bridged vertebrae, there was no association between syndesmophyte height and total strength (Figure 2 and Supplemental table 4). There was a significant inverse association between syndesmophyte height and trabecular strength, and a significant direct association with the proportion cortical strength. There was no association with cortical strength.

Figure 2.

Association between total strength (top row), trabecular strength (second row), cortical strength (third row), and proportion cortical strength (bottom row) and syndesmophyte height, among non-bridged (left) and bridged (right) vertebrae, among 119 vertebrae in the finite element analysis.

Among bridged vertebrae, there were strong inverse associations with total strength, trabecular strength, and cortical strength, and a direct association with proportion cortical strength (Figure 2 and Supplemental table 4). Results of trabecular strength were therefore consistent with those of tBMD, indicating more syndesmophytes in vertebrae with lower trabecular density and strength. The finite element analysis provided the unique information that syndesmophytes were more common among vertebrae that had higher cortical strength relative to total strength.

Longitudinal associations

Of 165 vertebrae potentially evaluable from the 33 patients in the longitudinal study, we included data on 149 vertebrae. Data were missing for 11 vertebrae because of limited field of view on either the baseline or year 2 scan, and the year 2 scan of one patient could not be processed for tBMD. Over the 2 years, tBMD decreased by a mean (± standard deviation) of 4.6% ± 9.2% in all vertebrae, 3.6% ± 7.6% in non-bridged vertebrae, and 5.9% ± 10.8% in bridged vertebrae. Syndesmophyte height increased by a mean of 2.7 ± 4.8 units. Based on mean ASDAS, 18%, 30%, 46%, and 6% of patients had inactive disease, moderate activity, high activity, and very high activity, respectively.

We first tested whether baseline tBMD was associated with changes in syndesmophyte height over two years (Figure 3 and Supplemental table 5). Among the 83 non-bridged vertebrae, vertebrae with lower baseline tBMD had more syndesmophyte growth over the subsequent two years (adjusted beta −0.01; 95% CI −0.019, −0.0012)). Because 50 vertebrae had no syndesmophytes at baseline, and because new syndesmophytes rarely develop in two years, we repeated the analysis among the 33 vertebrae with syndesmophytes at baseline. Results were similar in this subgroup, with a slightly larger effect estimate (adjusted beta −0.0133; 95% CI −0.024, −0.0024). In contrast, among bridged vertebrae, there was no association between tBMD at baseline and subsequent syndesmophyte growth (Supplemental table 5). Mean ASDAS was not associated with change in syndesmophytes in these models.

Figure 3.

Association between trabecular bone mineral density (tBMD) (top) and vertebral strength (bottom) at baseline and change in syndesmophyte height over two years among non-bridged and bridged vertebrae. The Y axis scales are different to accommodate the larger changes in syndesmophytes among bridged vertebrae.

We then tested whether baseline vertebral strength was associated with changes in syndesmophyte height over two years. Among non-bridged vertebrae, lower vertebral total strength at baseline was associated with more syndesmophyte growth over the subsequent two years (adjusted beta −0.0003; 95% CI −0.0004, −0.0002) (Figure 3 and Supplemental table 5). This association was also present for trabecular strength and cortical strength separately. In contrast, among bridged vertebrae, there were no associations between vertebral strength and syndesmophyte growth (Supplemental table 5).

We repeated these analyses after excluding 11 vertebrae from the longitudinal cohort, and 9 vertebrae from the finite element analysis cohort with adjacent degenerated discs. Results were very similar to the overall cohorts, with low tBMD and low vertebral strength predictive of more syndesmophyte growth only in non-bridged vertebrae (Supplemental table 6).

We next tested whether baseline syndesmophyte height was associated with the percent change in tBMD over the subsequent two years (Figure 4 and Supplemental table 7). There was no association among non-bridged vertebrae. Among bridged vertebrae, those with higher syndesmophyte height at baseline had greater losses of tBMD over two years (adjusted beta −0.001; 95% CI −0.0017, −0.0004). Results were similar after excluding vertebrae with adjacent degenerated discs (Supplemental table 8).

Figure 4.

Association between syndesmophytes at baseline and percent change in trabecular bone mineral density (tBMD) at two years among non-bridged and bridged vertebrae. The X axis scales are different to accommodate the range of syndesmophytes present among bridged vertebrae.

DISCUSSION

The presence of vertebral bone loss and syndesmophyte formation in close proximity in AS represents a major paradox, and the relationship between these processes has not been clearly defined.[25] By examining tBMD, bone strength, and syndesmophytes in the same vertebra using CT, our results demonstrate that this association differs depending on the presence of bridging. Among non-bridged vertebrae, those with lower tBMD and lower bone strength tended to have more syndesmophytes at baseline, and had more syndesmophyte growth over time. However, among bridged vertebrae, there was a strong inverse association between tBMD and syndesmophytes in the cross-sectional analysis, particularly in vertebrae with more extensive bridging. In the longitudinal analysis, more extensively bridged vertebrae experienced greater loss of tBMD over time, while loss of tBMD did not consistently occur among vertebrae without bridging. Therefore, bridging plays a critical role in modifying the process of vertebral bone loss in AS.

Previous cross-sectional studies that used QCT reported that lower lumbar tBMD was associated with the presence of syndesmophytes,[24,25] more extensive spinal changes,[23] or bridging,[10,22] but the proliferative changes in these studies were not restricted to the lumbar spine. Only Karberg et al [24] limited the examination of syndesmophytes to the lumbar spine, and only two studies adjusted for patient age.[23,25] In a ten-year longitudinal study of 15 patients, mean lumbar QCT-BMD was lower at baseline, and decreased more over time, among patients with any bridged vertebrae than patients without bridging.[21] These results suggested that bridging may have a more specific association with vertebral bone loss than syndesmophytes in general.

Our longitudinal results indicated that, in the absence of bridging, and presumably earlier in the course of AS, lower vertebral tBMD and lower bone strength was associated with greater subsequent syndesmophyte growth. In early AS, vertebral osteopenia has been attributed to the effects of both local and systemic inflammation.[35–38] This hypothesis is supported by improvement in vertebral BMD with TNFi treatment.[39–41] We did not find consistent associations between mean ASDAS and changes in tBMD or syndesmophytes over time, possibly because tBMD mediated the association between inflammation and syndesmophyte growth. Alternatively, the lack of consistent associations may be because the ASDAS was not assessed frequently enough, because of limitations in this measures, or because systemic inflammation may be less important than local inflammation in mediating these changes.[37,38] However, higher ASDAS as well as lower vertebral strength were associated with more syndesmophyte growth over time among non-bridged vertebrae. Current TNFi use was not associated with tBMD in our analysis, but it is important to note that this does not reflect changes that might follow the initiation of TNFi treatment.

Consistent with Wolff’s Law and concepts of bone adaptation, vertebral tBMD and strength are normally maintained by functional forces acting on the spine, explaining why vertebral bone is rapidly lost during space flight or prolonged bedrest.[42–45] In healthy individuals, age-related decreases in vertebral BMD affect trabecular bone primarily, particularly in men, such that cortical bone provides a larger proportion of the resistance to compressive forces in older persons.[46–49]. Consistent with this pattern, results of the finite element analysis indicated that lower trabecular strength in AS was associated with a larger relative contribution of cortical strength to total vertebral strength. Our results suggest a model in which, in the absence of bridging (perhaps more common in early AS), low tBMD, lower vertebral strength, and a higher proportional contribution of cortical strength to total bone strength promote syndesmophyte development (Figure 5). In this model, inflammation may lead to syndesmophyte development directly via enthesitis, but also indirectly through low tBMD and lower vertebral strength.[13] Loss of tBMD may also explain the historical observation of rapid spine fusion in patients with AS who were treated with body casts.[50,51] That vertebral strength had stronger associations with syndesmophyte growth than did tBMD may indicate the importance of changes in bone microarchitecture, as previously reported to be present in peripheral bones in AS.[52]

Figure 5.

Proposed model of the associations between vertebral bone loss and syndesmophyte development. In the absence of bridging (left side), inflammation at entheses directly leads to new syndesmophyte development while local and/or systemic inflammation leads to loss of vertebral trabecular bone mineral density (BMD). Low tBMD contributes to syndesmophyte development (arrow) as a result of transfer of stresses from trabecular bone to cortical bone. Over time, in vertebrae that develop bridging (right side), the additional structural support provided by this column of bone serves to off-load functional forces from trabecular bone and results in further loss of tBMD. In turn, the loss of tBMD may promote more extensive bridging as more functional stress is shifted from trabecular to cortical bone.

Among bridged vertebrae, our data indicate that the converse association holds: more extensive bridging is associated with lower tBMD and trabecular strength, and leads to greater loss of tBMD over time. Although this association may reflect the effects of vertebral inflammation, the strong association with the extent of bridging is more consistent with a mechanical effect. Specifically, trabecular bone loss in the setting of extensive bridging may be a consequence of vertebral off-loading. Extensive bridging may act as a scaffold (or column, in engineering terms) that relieves compressive forces from trabecular bone, and transfers these forces to the cortical shells of the adjoined vertebrae (Figure 5). Experiencing less compressive force, trabecular bone is lost more rapidly. This may also establish a positive feedback loop between low vertebral tBMD and bridging. Additional longitudinal studies with longer follow-up are needed to confirm this hypothesis.

A major strength of this study was the use of CT to quantitate precisely syndesmophytes and volumetric tBMD, thereby avoiding the confounding found in many prior studies. Additionally, we examined associations within individual vertebrae, and stratified by bridging. Finite element analysis provided biomechanical evidence of links between tBMD, vertebral strength and syndesmophyte formation. Our study is limited in that we studied relatively few patients longitudinally, and for only 2 years. We do not know if similar associations are present in other spine regions. More patients and larger vertebral segments are needed to study whether bridging across several vertebrae differentially affects bone loss. The associations we found were not related to disc degeneration, but further study of the potential role of disc degeneration in syndesmophyte development is needed.

This study indicates that the association between vertebral bone loss and syndesmophytes in AS is complex, likely stage-specific, and bidirectional. Although previously suspected, clear evidence of the unique role of bridging in mediating low tBMD has been lacking. Low tBMD in AS should therefore be considered in the context of whether bridging syndesmophytes are present or not, as the primary mechanisms underlying low bone density are likely different. Treatments targeting inflammation may be more effective in improving tBMD in non-bridged vertebrae. Conversely, among bridged vertebrae, where mechanical off-loading may be primarily responsible, antiresorptive or anabolic medications may be more effective treatments to improve tBMD.