Ankylosing spondylitis (AS) is a chronic inflammatory rheumatic disease, mainly affecting the sacroiliacal joints, vertebrae and intervertebral discs, leading to syndesmophyte formation and impaired back mobility. In AS two enhanced but opposite bone remodelling processes are taking place in close vicinity within the spine; these are pathologic new bone formation in the cortical zone of the vertebrae, the zygapophyseal joints, and the ligamentous apparatus and excessive loss of trabecular bone in the centre of the vertebral body leading to osteoporosis. Earlier studies have demonstrated an increased prevalence of osteoporosis and significantly lower bone mineral density (BMD) in AS patients compared with sex and age matched controls [1, 2]. Osteoporosis has also been shown to be present in mild AS and in early disease [2, 3]. AS patients have a high risk of vertebral fractures, which can be complicated by neurological injuries [4, 5].

In advanced AS it can be difficult to interpret lumbar spine BMD measured with dual-energy x-ray absorptiometry (DXA) in the anteroposterior (AP) projection. The new bone formation that is characteristic of AS causes an overestimation of the total BMD and values can be normal or high, even when osteoporosis is present. Spinal hyperostosis in AS is frequently located around the zygapophyseal joints, the vertebral endplates and the annulus fibrosus of the discs, and to a lesser extent, on the lateral sides [6]. Lateral DXA scanning of the lumbar spine allows exclusive examination of the vertebral body, which consists of 80% trabecular bone, and excludes the zygapophyseal joints, endplates and both anterior and posterior syndesmophytes from the measurement. Consequently, lateral scanning could be a way to reduce the problem of overestimation of lumbar spine BMD in AS. With older single-beam DXA systems patients were examined lying on their side and the precision was low. Modern lateral scans with the patient positioned supine offer precision similar to regular AP spine DXA scans [7].

The combination of AP and lateral DXA allows assessment of three-dimensional volumetric BMD (vBMD = bone mineral content/volume). Areal BMD (aBMD) is measured two-dimensionally, without taking the size of bone in the third dimension into account. Consequently, aBMD depends on both bone density and bone size, whereas vBMD is independent of bone size. Men have larger bones than women and consequently have higher aBMD, but vBMD is equal in both sexes [8, 9].

BMD varies between different populations. Sweden has a high prevalence of osteoporosis and osteoporosis-related fractures [10–12]. Although studies of osteoporosis in AS have been performed in other countries, this is to our knowledge the first study in Scandinavia. The aims of the present study were to investigate the prevalence of, and risk factors for osteoporosis in patients with AS, to compare lumbar spine BMD measured in the AP and lateral projection, including estimated vBMD, and to study how these measures change with progressive ankylosis.

Of the 538 AS patients registered in the hospitals' databases, 177 patients did not meet the modified New York criteria for AS or had exclusion criteria. Out of a total of 361 patients invited to participate, 72 declined, 60 did not respond to the invitation and 19 did not meet the inclusion criteria. Of the 210 patients included, 6 did not attend the DXA or radiography appointments and were therefore excluded. The patients who fulfilled the inclusion criteria but declined to participate or did not respond to the invitation (n = 151) were significantly younger than the patients included in the study (46 ± 13 years vs. 50 ± 13 years; P = 0.007), but the sex distribution was the same among patients who were or were not included.

Prevalence of reduced BMD

Among patients under 50 years of age (n = 103), 35 patients (34%) had a BMD Z-score < -1.0 at the hip and/or AP lumbar spine and 5 patients (4.9%) had BMD below the expected range for age, that is, a Z-score ≤ -2.0. Among patients 50 years of age or older (n = 101), 44 (43.6%) had osteopenia and 21 (20.8%) had osteoporosis using the WHO definition [22]. The lumbar spine was the most common location for osteoporosis or BMD below the expected range for age (10%), followed by radius (8%) and femoral neck (5%). The prevalence of reduced BMD in different locations is shown in Table 2.

Measurement site	Age-group, years	BMD, mean ± SD	T-score, mean ± SD	Z-score, mean ± SD	Patients with osteoporosis/BMD below expected range for age, number (%)a	Patients with osteopenia, number (%)	Patients with normal BMD number (%)
Women
AP lumbar spine	< 50	1.014 ± 0.134	-0.30 ± 1.22	0.02 ± 1.25	2 (5)	0 (0)	41 (95)
	≥ 50	0.899 ± 0.150	-1.35 ± 1.37	0.14 ± 1.38	12 (27)	15 (34)	17 (39)
Lat lumbar spine	< 50	0.741 ± 0.102	-0.94 ± 1.22	-0.25 ± 1.26	3 (7)	0 (0)	40 (93)
	≥ 50	0.616 ± 0.114	-2.43 ± 1.36	-0.05 ± 1.42	20 (45)	15 (34)	9 (20)
Total hip	< 50	0.925 ± 0.122	-0.13 ± 1.00	0.07 ± 0.99	0 (0)	0 (0)	43 (100)
	≥ 50	0.832 ± 0.108	-0.89 ± 0.88	0.10 ± 0.85	0 (0)	20 (46)	23 (54)
Femoral neck	< 50	0.804 ± 0.127	-0.41 ± 1.15	-0.06 ± 1.13	0(0)	0 (0)	43 (100)
	≥ 50	0.702 ± 0.103	-1.32 ± 0.93	-0.01 ± 0.91	4 (9)	24 (56)	15 (35)
Total radius	< 50	0.576 ± 0.038	-0.05 ± 0.70	0.32 ± 0.74	0(0)	0 (0)	43 (100)
	≥ 50	0.517 ± 0.065	-1.14 ± 1.21	0.18 ± 1.09	7 (16)	17 (39)	20 (45)
Men
AP lumbar spine	< 50	1.057 ± 0.154	-0.31 ± 1.41	-0.18 ± 1.41	3 (5)	0 (0)	57 (95)
	≥ 50	1.092 ± 0.184	0.18 ± 1.67	0.63 ± 1.72	4 (7)	8 (14)	44 (79)
Lat lumbar spine	< 50	0.782 ± 0.131			NA	NA	NA
	≥ 50	0.740 ± 0.132			NA	NA	NA
Total hip	< 50	0.999 ± 0.137	-0.23 ± 0.92	-0.05 ± 0.90	1 (2)	0 (0)	59 (98)
	≥ 50	0.963 ± 0.136	-0.46 ± 0.90	0.00 ± 0.93	0 (0)	19 (33)	38 (67)
Femoral neck	< 50	0.842 ± 0.130	-0.67 ± 0.95	-0.21 ± 0.93	1 (2)	0 (0)	59 (98)
	≥ 50	0.773 ± 0.122	-1.16 ± 0.90	-0.21 ± 0.90	5 (9)	25 (44)	27 (47)
Total radius	< 50	0.668 ± 0.050	-0.34 ± 0.93	-0.12 ± 0.91	1 (2)	0 (0)	59 (98)
	≥ 50	0.633 ± 0.063	-1.03 ± 1.21	-0.25 ± 1.20	9 (16)	18 (33)	28 (51)

Definition of reduced BMD in patients younger than 50 years, BMD below expected range for age (Z-score ≤ -2.0 SD), BMD within expected range for age (Z-score > -2.0 SD); in patients 50 years or older, World Health Organization definition of osteoporosis: T-score ≤ - 2.5 SD, osteopenia: T-score < -1 to > -2.5 SD; normal BMD, T-score ≥ -1SD. For lateral lumbar spine BMD in men there are no reference values available for calculation of T- or Z-scores. ^aData presented as the sum of patients younger than 50 years, with a Z-score ≤ -2.0 SD and patients 50 years or older with a T-score ≤ -2.5 SD.

AP, anteroposterior; Lat, lateral; BMD, bone mineral density.

Men had significantly lower BMD at the femoral neck (mean Z-score -0.215, P = 0.012), compared with the age- and sex-matched reference values in the DXA-scanner software. Mean Z-scores were negative at all measurement sites except the lumbar spine (AP projection), but was not significantly different to zero. Significantly more women aged 50 years or older had osteoporosis on measurement of AP lumbar spine than men (12/44, 30% vs. 4/56, 7%, P < 0.001). Consequently, diagnosis of osteoporosis and osteopenia using the WHO definition was more prevalent among women (13/44, 30% and 21/44, 48%, respectively) compared with men (8/56, 14% and 23/56, 41%, respectively; P = 0.042). BMD below the expected range for age was equally common in women and men below 50. Men had a significantly higher mSASSS score (median 8, range 0 to 72) than women (median 2, range 0 to 46) (P < 0.001) (Figure 1). Men had significantly higher aBMD at all measurement sites (P < 0.001 for every location), but for lumbar vBMD there was no significant difference between the sexes (P = 0.36). The BASDAI, BASFI, BASMI and ASDAS were evenly distributed among women and men.

A total of 36% of women and 8% of men had ever undergone a DXA examination before the study. Eight patients were on bisphosphonates at the inclusion. After the study new treatment with calcium, vitamin D and bisphosphonates, according to national guidelines, was initiated in 30 patients (14.2%) and calcium and vitamin D alone was initiated in 28 patients (13.2%).

Comparison of anteroposterior DXA and lateral DXA

Significantly more women had a lumbar spine BMD T-score ≤ -2.5 (age ≥ 50) or Z-score ≤ -2.0 (age < 50) when measured by lateral DXA (n = 23, 26%) compared with AP DXA (n = 14, 16%; P = < 0.001) (Table 2). T- and Z-scores for lateral spine DXA were not available for the men.

Lumbar spine BMD was significantly higher measured in the AP compared with the lateral projection (P < 0.001) (Table 1). The difference (AP minus lateral projection DXA BMD) increased with increasing mSASSS (r_S = 0.389, P < 0.001), BASMI (r_S = 0.296, P = 0.001), age (r_S = 0.309, P = 0.001) and disease duration (r_S = 0.268, P = 0.004) in men, but not in women.

Since men had significantly higher mSASSS than women the male subgroup was chosen to further study the effects of increasing ankylosis on BMD. Increasing mSASSS in men was significantly correlated with lower lumbar spine vBMD (r_S = -0.389, P < 0.001), lower lateral spine BMD (r_S = -0.191, P = 0.041), femoral neck BMD (r_S = -0.324, P < 0.001) total hip BMD (r_S = -0.201, P = 0.030) and total radial BMD (r_S = -0.269, P = 0.004), but not with AP lumbar spine BMD (r_S = 0.152, P = 0.103) (Figure 2). Higher BASMI in men also correlated significantly with lower BMD at all measurement sites, except for AP lumbar spine BMD, which tended to be higher, but was not significantly so.

The present study demonstrates that osteoporosis is common in Swedish patients with AS, but is often undiagnosed and thus untreated. BMD was below the expected range for age in 5% of patients under the age of 50 years. Osteoporosis, as defined by the WHO, was diagnosed in 21% of patients aged 50 years or older (in 30% of the women and 14% of the men). Most patients with osteoporosis were undiagnosed before the study. The prevalence of osteoporosis found in the present study is comparable with the prevalence in the Swedish general population aged 50 to 80 years, where 21.2% of women and 6.3% of men have been identified as having osteoporosis [25]. The results indicate that the increased risk of osteoporosis in AS compared with the general population is especially accentuated in male patients, which is in accordance with findings in another study [26].

In the current study osteoporosis and osteopenia in the lumbar spine were more common in women than in men in the age-group of 50 years or older when measured with AP lumbar DXA, and thus more women were diagnosed as osteoporotic. The prevalence of osteoporosis and osteopenia was equal among women and men at all other locations. The high prevalence of syndesmophytes in the men may have resulted in falsely elevated BMD causing an underestimation of male osteoporosis at the lumbar spine.

The results from the present study indicate that osteoporosis and osteopenia affect both the central and the peripheral skeleton. We found almost as many patients with BMD below the expected range for age or osteoporosis at the radius (n = 17, 8%) as in the lumbar spine (n = 24, 12%). Results from earlier studies indicate that osteoporosis in AS is predominantly confined to the central skeleton [27, 28]. In one study no correlation was found between bone density at the forearm measured with peripheral quantitative computed tomography (pQCT) and DXA and quantitative CT (QCT) measurements of BMD at the lumbar spine or hip [29]. Quantitative ultrasound studies of the heel in patients with AS have inconsistent findings, with normal results in one study, and signs of peripheral osteoporosis in another [30, 31]. The theory that osteoporosis is a general process affecting the whole skeleton in AS was supported by a study of bone biopsies from the iliac crest, showing trabecular thinning and low trabecular peripheral bone volume strongly correlated with lumbar spine BMD measured using QCT [32].

There is uncertainty about how to treat osteoporosis in patients with AS. Bisphosphonates have been studied in respect to their effect on disease activity in AS, but their effects on fractures, BMD and the new bone formation in AS needs to be further investigated. Pamidronate has been reported to hamper disease activity in AS [33, 34]. In a recent placebo-controlled study of alendronate 70 mg weekly, no improvement of AS symptoms or disease activity was found [35].

In the current study we found that low BMD was associated with older age, longstanding disease, syndesmophyte formation, impaired back mobility, history of coxitis, use of glucocorticoids and laboratory signs of inflammation. Menopause was a strong risk factor for women. No connection was found between low BMD and the disease indices BASDAI, BASFI, BAS-G or ASDAS. BASMI and ASDAS were associated with inflammatory parameters (ESR, CRP), but BASDAI, BASFI and BAS-G were not.

The association between extensive syndesmophyte formation, restriction of spinal movement and osteoporosis has been demonstrated previously [26, 29, 36]. One study found significant correlation between low lumbar spine BMD and elevated ESR and CRP [37]. Two follow-up studies have shown that patients with AS and persistent inflammation, that is, with elevated ESR or CRP, developed significant decreases in BMD, whereas patients with low inflammatory activity did not [38, 39].

In men, who had significantly higher mSASSS than women, increasing mSASSS and BASMI were significantly associated with decreasing vBMD and lateral BMD at the lumbar spine, along with lower BMD at the hip and radius, while AP lumbar BMD had a non-significant tendency to increase. Our interpretation of the results is that in comparison with AP spine BMD, lumbar spine vBMD and lateral lumbar spine BMD are less affected by the new bone formation in AS and hence are more adequate tools in assessing osteoporosis and osteopenia, especially in male patients in AS.

Lateral lumbar spine BMD is usually lower than AP BMD, because the lateral DXA scan measures only the trabecular-rich vertebral body, whereas the AP scan includes both the vertebral body and the posterior part of the vertebra, mainly consisting of dense cortical bone. AP scanning is also affected by artefacts due to to osteophytes, aortic calcifications and degenerative changes in the facet joints of elderly people and from hyperostosis in AS. The trabecular bone is more metabolically active than the cortical bone, therefore a decrease in BMD is first found in the trabecular bone. Consequently lateral lumbar spine BMD declines faster than AP BMD in early osteoporosis [40, 41].

The current study demonstrates that lateral lumbar spine DXA is more sensitive than AP DXA in detecting osteoporosis and osteopenia in AS. The same results have been reported in two studies using lateral DXA of vertebra L3 in patients with AS [42, 43]. Other studies have shown that lateral DXA is more sensitive than AP DXA in detecting osteopenia and osteoporosis in degenerative spinal disease [44]. In one study of 100 AS patients and 58 healthy controls assessed with both AP and lateral lumbar DXA using a scanner similar to the one used in the current study, the authors reported that lumbar spine BMD was significantly lower in AS patients compared with healthy controls when measured by lateral projection DXA, but not when measured by AP DXA [45]. However, to apply lateral DXA in clinical practice, reference intervals based on measurements on large populations of healthy men and women are required. Most likely, new threshold values for definition of osteoporosis have to be defined to avoid overestimation of osteoporosis with lateral DXA. The current WHO definition of osteoporosis is based on the PA projection and according to the Official Positions of the International Society for Clinical Densitometry 2007, the lateral spine should not be used for diagnosis of osteoporosis, but it may have a role in monitoring [23].

Osteoporosis and osteopenia are common in Swedish patients with AS and affected half of our study population. Low BMD was found in both the central and the peripheral skeleton. Osteoporosis was often undiagnosed and untreated, particularly in male patients with AS. Older age and high disease burden, that is, long disease duration, impaired back mobility, syndesmophyte formation and elevated inflammatory parameters, indicated increased risk of osteoporosis. Lateral and vBMD at the lumbar spine were less affected by progressive ankylosis in AS compared with AP BMD. In addition, lateral DXA was more sensitive in detecting osteoporosis and osteopenia than AP DXA. We suggest that lateral lumbar spine DXA with calculation of vBMD may become valuable tools in the diagnosis and follow-up of osteoporosis in AS, but validation of the methods in larger reference populations is needed.