Methods

The data in this study are derived from the General Practice Research Database (GPRD).15 This database comprises the entire computerised medical records of a sample of general practitioners in the UK. General practitioners play a key role in the healthcare system because they are responsible for primary health care and specialist referrals. The database comprises centralised information not only from the general practitioners themselves but also from specialist referrals and hospital attendances. This study included 683 practices currently incorporated in the GRPD and thereby comprises a 6% sample of the UK population. The data recorded in the GRPD include demographic information, prescription details, clinical events, preventive care provided, specialist referrals, hospital admissions and major outcomes. Clinical data are stored and retrieved by means of Oxford Medical Information Systems and Read codes for diseases that are cross-referenced to the International Classification of Diseases (ICD-9). Data collection for the GRPD began in 1988 and, for this study, ended in 1999.

A case-control study was conducted using information from the GRPD. Cases were registered patients aged 18 years and older who had any clinical fracture in their medical records (vertebral (wedge, crush or diabolo) or non-vertebral). Control patients were adults without a history of fracture in their records. They were matched to case patients by age (within 1 year) and sex. A history of AS was assessed from the medical records (according to the modified New York criteria).16 The number of patients with AS was examined in both cases and controls. Prescriptions for non-steroidal anti-inflammatory drugs (NSAIDs) and disease-modifying antirheumatic drugs (DMARDs) in the 6 months before the fracture date in patients or corresponding date for the controls were also evaluated. Prescribing of bone-protective agents including bisphosphonates, hormone replacement therapy and vitamin D in the previous 6 months was also examined. A medical history of diseases associated with AS such as inflammatory bowel disease (IBD) and psoriasis, as well as a history of back pain and joint replacement, were also assessed from the medical records.

Statistical analysis

χ² tests were used to test differences in proportions. Conditional logistic regression was used to estimate odds ratios (OR) for the risk of fracture in patients with AS. The analysis was controlled for a wide range of clinical variables and medication use possibly associated with the risk of fracture including history of heart failure, cerebrovascular disease, diabetes mellitus, rheumatoid arthritis, psychotic disorder, depression, thyrotoxicosis, seizure and chronic obstructive pulmonary disease. Prescriptions for thiazides, hormone replacement therapy, NSAIDs, hypnotics/anxiolytics, antipsychotics, antidepressants, antiepileptics, anti-Parkinson drugs, all inhaled corticosteroids and bronchodilators in the 6 months before the fracture were also considered potential confounding variables. Smoking history and body mass index were included if entered in the database. Because smoking history and body mass index were not part of the standard data collection, this information was missing for about half of the patients.

Results

A total of 231 436 cases of fracture were selected and matched with 231 362 controls. The characteristics of the cases and controls are shown in table 1. The groups were well matched for age and sex, but there were significant differences in the number of potential confounders. The patients with fracture used more medication, especially NSAIDs (13.9% vs 9.1%), but also sulfasalazine (0.24% vs 0.19%), corticosteroids (4.1% vs 2.3%) and hypnotics/anxiolytics (9.6% vs 6.2%), and had a higher number of relevant concomitant diseases such as IBD (0.5% vs 0.4%) and psoriasis (2.7% vs 2.0%). More cases with fracture reported back pain in the preceding 6 months. AS was diagnosed in 758 patients, 416 (0.18%) of the patients with fracture and 342 (0.15%) of the controls (p<0.05). The age, sex, use of medications and relevant co-morbidities for the patients with AS, with and without fractures, are shown in table 2. Patients with AS with fractures tended to be somewhat older than patients with AS without fractures (p = 0.09) because there were fewer patients aged <30 years in the fracture group. This group also used systemic corticosteroids more frequently (8.4% vs 4.1%), were more likely to have IBD (4.8% vs 1.8%) and suffered more frequently from back pain (17.3% vs 12%).

Table 3 provides the OR for any clinical fracture, as well as for clinical vertebral and non-vertebral fractures separately. Patients with AS had an increased crude risk of clinical vertebral fractures (OR 3.08; 95% CI 1.56 to 6.08) that remained significantly increased after adjustment for potentially confounding factors (OR 3.26; 95% CI 1.51 to 7.02) compared with those without AS. The risk of forearm and hip fracture was not significantly increased (OR 1.21; 95% CI 0.87 to 1.69 and OR 0.77; 95% CI 0.43 to 1.37, respectively).

In table 4 the risk of fracture has been stratified for relevant concomitant diseases and medications in patients with AS. The adjusted risk of any clinical fracture was increased in patients with AS with a history of IBD (OR 2.79; 95% CI 1.10 to 7.08) but was decreased in patients with AS taking NSAIDs (OR 0.65; 95% CI 0.50 to 0.84), although this protective effect became evident only after adjustment for confounding variables. The risk for any fracture was not associated with back pain, psoriasis, joint replacement therapy or use of sulfasalazine.

Discussion

In this study we have shown that patients with a diagnosis of AS have an increased risk of clinical vertebral fractures without an increased risk of non-vertebral fractures. This is in agreement with reports in the literature that, in AS, an increased risk of fractures is limited to the axial skeleton.4 Although the patients with AS tended to be older than the control patients, the increased risk remained after adjustment for age. Similarly, as shown in the analysis, other variables did not significantly confound the observed effect, which adds to the validity of the findings. Interestingly, the risk of any clinical fracture was increased in patients with AS with concomitant IBD. In contrast, the risk of any clinical fracture was decreased in patients with AS treated with NSAIDs.

This primary care-based nested case-control study confirms earlier findings in smaller groups of patients with AS. Cooper et al found an increased risk of spinal deformities (compression, wedging or collapse of one or more thoracic or lumbar vertebral bodies) compared with the normal population. We have described 59 patients with AS who reported clinical vertebral fractures early in the disease, mainly without or after minimal trauma.6 These data underline the greater vulnerability of the spine in patients with AS with respect to clinical vertebral fractures. As mentioned previously, there is increasing evidence to support the view that changes in the biomechanical properties of the spine such as osteoporosis (which renders the vertebrae prone to fracture) and osteoproliferation (which makes the spine less flexible) may play an important intermediate role. We have reported earlier that wedged vertebral fractures are related to radiographic signs of osteoproliferation, indicating that—at least in part—an increased risk of fracture could be associated with more extensive disease.12 The observation that patients using systemic glucocorticoids and those with recent back pain had more fractures underpins the hypothesis of an increased risk of fractures in inflammatory disease, as has been shown in patients with rheumatoid arthritis.17 While patients with AS have an increased fracture risk limited to the spine, those with AS and associated IBD have an increased risk of both vertebral and non-vertebral fractures. Schoon et al identified several clinical risk factors for low bone mass density (BMD) in patients with IBD, especially Crohn’s disease.18 They also showed that patients with recent onset of IBD had normal BMD, suggesting that low BMD develops in the course of the disease process.19 Heijckmann et al reported a high prevalence of vertebral deformities in Crohn’s disease and ulcerative colitis, especially in older patients. This was associated with higher glucocorticoid use.20

We used the GPRD database to select our patients. The validity of the results is directly associated with the quality of the data. The data quality of each entry into the GRPD is measured against specific targets to ensure research standards are met. Only data from practices that pass this quality control are compiled to form the GRPD database.21 22 A problem in a cohort study could be that previous selection may not be ruled out. By taking a very large sample, we can assume that possible bias is equally distributed in both groups. Theoretically, our results may have been biased by an inappropriate classification of AS. An argument that adds to the validity of our results is that we found a positive relationship between a diagnosis of AS and the presence of a clinical fracture that was entirely in line with our primary hypothesis. Any relevant misclassification of AS would have resulted in a lack of association rather than in the spurious opposite. Furthermore, we incorporated as many confounders known in osteoporosis and AS research as possible to be able to adjust for these factors in the statistical analysis. A broad spectrum of medication and concomitant diseases was taken into account. These items seemed to be equally divided in both groups. Owing to the very large sample (nearly half a million), small proportional differences between groups could have had a large influence on the results, especially if this happened preferably in one group (eg, patients with AS with fractures). Fortunately, the patients with AS with and without fractures were comparable in most respects.

Taken together, these data point to a dose-response effect of chronic inflammatory diseases with respect to the risk of fracture. AS conveys a risk of vertebral fractures. Concomitant Crohn’s disease—which is considered a more systemic disease and is more frequently treated with systemic corticosteroids—adds to this moderate risk which is reflected by a higher fracture rate for vertebral and non-vertebral fractures. Another interesting observation is the protective effect of NSAID use on the risk of fracture. It is well known that NSAIDs reduce the risk of heterotopic bone formation after hip arthroplasty.23 In one study, continuous use of NSAIDs decreased the formation of syndesmophytes, thus preventing progressive rigidity of the spine with decreased vulnerability as a consequence.24 The influence of NSAIDs on fracture repair in mouse models has been described.25 26 It is suggested that the influence of NSAIDs on bone metabolism is dose-dependent. Unfortunately, we were not able to investigate a dose-response relation in our group of patients using NSAIDs due to insufficient information on the doses used. One other possible explanation could be that relief of inflammatory back pain and stiffness, as well as the resulting improved physical function due to NSAID use, may help in maintaining bone mass by improved physical activity. Another positive effect of improved physical condition could be a reduced risk of falling. Data suggest that a programme of muscle strengthening and balance training decreases the risk of a fall, and the use of NSAIDs could have a positive effect.27 28

In contrast to rheumatoid arthritis, the pharmaceutical treatment of osteoporosis in patients with AS is not yet common practice. There are few data supporting the efficacy of osteoporosis treatment in AS. Our findings and those of others point to a role of osteoporosis—possibly in combination with osteoproliferation—in the occurrence of vertebral fractures in AS. This study highlights the need to develop strategies to identify patients with AS with a risk of vertebral fractures, both in terms of osteoporosis and excessive osteoproliferation. Appropriate treatment intervention studies can be designed. The mechanism by which NSAID use is associated with a reduced risk of clinical fractures warrants further research.