The Assessment of SpondyloArthritis international Society (ASAS) classification criteria for axial spondyloarthritis (SpA) allow SpA classification of HLA–B27–positive patients if ≥2 specific clinical SpA features are present. We investigated the performance of these clinical ASAS criteria in a population with a high prevalence of HLA–B27.
A total of 807 persons reporting chronic back pain (CBP) lasting for >4 weeks during a population survey underwent a clinical, laboratory, and radiologic evaluation. The ASAS criteria for axial SpA were then used to determine classification status.
Only 332 patients (41% of all CBP patients) fulfilled the prerequisite ASAS definitions for CBP (duration of ≥3 months and onset at age <45 years). In this ASAS-defined CBP cohort (51% women, CBP onset at age 27.2 years, 17% HLA–B27 positive), ASAS classification criteria for axial SpA were met by 8.4% of patients. Radiographic SpA by the modified New York criteria was present in 2.4%, while 6% fulfilled the clinical arm of the ASAS SpA criteria only. One-fifth of patients with clinical SpA developed radiographic evidence of SpA after a median of 8 years.
Application of the clinical ASAS classification criteria in an area with a high prevalence of HLA–B27 leads to significant increases in the prevalence of axial SpA compared to radiologic SpA among CBP patients. This increase in the prevalence of disease is likely to have significant ramifications for patient management and health care systems.
Ankylosing spondylitis (AS) is the prototypical disease within the spondyloarthritis (SpA) group, which further includes SpA associated with psoriasis, inflammatory bowel disease (IBD), reactive arthritis, and an undifferentiated form of SpA. All SpA share clinical manifestations such as inflammatory back pain (IBP), asymmetric arthritis of the lower extremity, enthesitis, dactylitis, uveitis, and a genetic background mediated by HLA–B27 (1). The modified New York classification criteria for AS (2) have often been applied as a diagnostic instrument by both clinicians and insurers, due to a lack of satisfying diagnostic criteria for AS and other SpA. The modified New York criteria are anchored around radiographic evidence of chronic arthritis in the sacroiliac (SI) joints, but these changes represent late and irreversible damage to the SI joint. With the modified New York criteria, diagnostic delays of 5 to 10 years have been common, and they are not helpful in the early phase of the disease (3, 4). New classification criteria for axial SpA have been developed and validated by the Assessment of SpondyloArthritis international Society (ASAS) study group in order to facilitate modern management of patients with SpA (3, 5). The ASAS classification criteria introduce magnetic resonance imaging (MRI) findings as a new radiographic criterion for SpA due to their ability to detect sacroiliitis and spinal enthesitis in an early phase of the disease, when definite changes cannot be seen on radiographs. There is evidence that bone marrow edema in the SI joints on MRI precedes the typical radiographic signs of chronic sacroiliitis, although not in all patients (6). Also, access to MRI is not universally available to patients due to costs or availability issues, and signs of inflammation on MRI may disappear spontaneously or with drug therapy (6).
The ASAS classification criteria contain a clinical arm where a positive result for HLA–B27 testing is the anchor criterion, which together with the presence of ≥2 specified clinical features allow SpA classification when radiograph or MRI findings are absent or equivocal. The rationale to include these patients in the axial SpA group is that they have comparable manifestations and overall disease activity to patients with definite (i.e., radiographically proven) SpA (7). The ASAS classification criteria for SpA were developed on the basis of data from various regions, including areas with a low prevalence of HLA–B27. Because it is unclear to which degree the population prevalence of HLA–B27 will influence the utility of the clinical arm of the ASAS classification criteria, we investigated its applicability in a population with a prevalence of HLA–B27 of 16%.
Significance & Innovations
The prevalence of spondyloarthritis (SpA) nearly triples with the use of the clinical arm of the Assessment of SpondyloArthritis international Society SpA criteria in a region with a high prevalence of HLA–B27.
Twenty percent of patients classified as having clinical SpA developed radiographic evidence of SpA.
Whether (subgroups of) patients with clinical SpA will benefit from biologic therapy needs further study, but based on these numbers this has potentially large ramifications for health care delivery.
PATIENTS AND METHODS
The University Hospital of Northern Norway is located in Tromsø, the largest municipality in Northern Norway with ∼65,000 inhabitants. The frequency of HLA–B27 in the region is 15.9% (8), the prevalence of AS is 0.4%, and the prevalence of psoriatic arthritis is 0.13% (4, 9). A government-sponsored population-based health survey, the Tromsø Health Survey, is carried out every 6 to 7 years to assess risk factors and the prevalence of several chronic conditions (10). This analysis concerns data from the Third Tromsø Study (1986–1987), where men and women in the age groups 20–61 and 20–56 years, respectively, were invited to take part. All persons who responded as having had low back pain for >4 weeks were invited to an outpatient visit at the rheumatology clinic, where a single physician obtained the following data through a standardized evaluation in all patients: demographic and socioeconomic data, detailed medical and family history, date at onset and localization of back and peripheral joint pain, drug use for musculoskeletal pain, and Health Assessment Questionnaire score. A complete physical examination was then performed with special attention to spinal mobility testing, swollen/tender joint counts, enthesitis, and psoriatic skin lesions. Blood samples were analyzed for levels of hemoglobin, white blood cells, erythrocyte sedimentation rate, and HLA–B27 status (by serologic testing). A radiographic examination of the total spine and SI joints was performed in consenting patients, and 2 experienced radiologists independently examined these films for the presence of SI changes defined in the modified New York criteria. All of these clinical data were gathered between 1989 and 1992, and thereafter were consecutively entered into an electronic database for later analysis. With 12 of 13 possible ASAS-defined variables for SpA classification available, we were able to classify HLA–B27–positive patients with 2 additional clinical SpA features to the clinical ASAS-defined SpA cohort (5, 11). While data on sacroiliitis by MRI were the missing variable, we had a complete set of radiographic data, allowing patients with evidence of sacroiliitis according to the modified New York criteria to be classified as having radiographic SpA. Finally, to investigate the predictive potential of the clinical ASAS classification criteria, we linked the survey database with our hospital diagnosis register in 2010 to determine the number of patients given a radiographically confirmed SpA diagnosis subsequent to the survey. Statistical analyses included analysis of variance testing for comparing means for parametric data and the Mann-Whitney U test and chi-square test for nonparametric data.
The initial survey visit was attended by 10,963 men and 10,863 women (78% response rate for men, 85% response rate for women), and a total of 807 patients (3.7% of the survey population) who reported chronic back pain (CBP) lasting for >4 weeks were investigated during a followup visit at our outpatient clinic. The prerequisites for applying the ASAS classification criteria (CBP lasting for ≥3 months and onset at age <45 years) (3) were present in 332 patients (41.1% of all CBP patients) (Figure 1). Compared to the remaining 475 patients (58.9%) that were considered to have nonspecific CBP, patients in the ASAS-defined CBP cohort were more likely to have IBP-defined ASAS, to present with ≥2 clinical SpA features combined, and to carry HLA–B27 (Table 1). Radiographic evidence of definite sacroiliitis was not seen in patients with nonspecific CBP.
Table 1. Demographic and clinical data at investigation for the ASAS-defined CBP study cohort of 332 persons compared to the control group of patients with unspecific CBP*
ASAS-defined CBP (n = 332)
Unspecified CBP (n = 475)
ASAS = Assessment of SpondyloArthritis international Society; CBP = chronic back pain; NS = not significant; ESR = erythrocyte sedimentation rate; IBP = inflammatory back pain; IBD = inflammatory bowel disease; SpA = spondyloarthritis.
Age at CBP onset, mean ± SD years
27.2 ± 9.9
26.5 ± 9.0
Age at study onset, mean ± SD years
40.8 ± 8.2
39.1 ± 8.7
Male sex, %
ESR ≥15 mm/hour, %
ASAS-defined IBP, %
≥2 SpA features, %
Sacroiliitis (by modified New York criteria), %
In the ASAS-defined CBP cohort, the ASAS classification criteria for axial SpA were met by 8.4% of patients (Figure 1 and Table 2). Radiographically confirmed axial SpA was present in 8 patients (2.4%), with modified New York criteria grade 2b lesions in 4 patients and modified New York criteria grade 3–4 lesions in 4 patients. Seven patients carried HLA–B27, whereas 1 patient was HLA–B27 negative but had complete bilateral obliteration of the SI joints (modified New York criteria grade 4). Spinal radiography additionally showed ankylosis in 2 of these patients, which was not observed in the other subgroups. Abnormal SI joint findings not meeting the modified New York criteria (grade 1, n = 5 and grade 2a lesions, n = 3) were present in another 8 patients. Four patients were subsequently classified as non-SpA and 4 were classified as clinical SpA.
Table 2. Comparison of clinical data by axial SpA classification status in the ASAS-defined CBP study cohort (n = 332)*
Values are the mean unless otherwise indicated. SpA = spondyloarthritis; ASAS = Assessment of SpondyloArthritis international Society; CBP = chronic back pain; ESR = erythrocyte sedimentation rate; IBP = inflammatory back pain; IBD = inflammatory bowel disease; NSAID = nonsteroidal antiinflammatory drug.
By chi-square test and Mann-Whitney U test results.
Age at CBP onset, years
Age at study onset, years
Male sex, %
HLA–B27 positive, %
IBP score (range 0–5)
NSAID response, %
Positive family history, %
No. of SpA features
Based on the clinical arm of the HLA–B27 presence together with ≥2 clinical SpA features, 20 patients (6%) were also classified as having axial SpA. Radiographic SI lesions were observed in 4 patients with clinical SpA, but did not meet the modified New York criteria (grade 1, n = 2 and grade 2a, n = 2).
When the 3 subgroups of ASAS-defined CBP were compared, there were significantly higher frequencies of most SpA features in the clinical and radiographic SpA groups, although not for IBD (Table 2). Peripheral arthritis was only seen in the radiographic SpA group, where male predominance was also more pronounced. We also analyzed the performance of individual ASAS SpA features in distinguishing SpA (n = 28) from non-SpA patients (n = 304) (Table 3). The highest diagnostic odds ratio (OR; [sensitivity × specificity]/[1 − sensitivity][1 − specificity]) was seen for HLA–B27 (diagnostic OR 326) and uveitis (diagnostic OR 16), but the diagnostic precision of these findings was 48% and 66%, respectively.
Table 3. Performance of individual features in distinguishing all SpA (n = 28) from non-SpA patients (n = 304) among patients with chronic back pain for ≥3 months and disease onset at age <45 years, with SpA defined according to the ASAS classification criteria*
ASAS SpA feature
SpA = spondyloarthritis; ASAS = Assessment of SpondyloArthritis international Society; LR = likelihood ratio; OR = odds ratio; PPV = positive predictive value; IBP = inflammatory back pain; NSAID = nonsteroidal antiinflammatory drug; ESR = erythrocyte sedimentation rate; IBD = inflammatory bowel disease; NC = not calculated.
Positive family history
A review of hospital records for patients in the ASAS-defined CBP study cohort not fulfilling the modified New York criteria at the time of the survey disclosed that 4 patients had, after a median of 8 years, developed radiographically confirmed SpA by radiography (n = 2) or computed tomography (n = 2). One additional patient had subsequent MRI findings compatible with sacroiliitis, although findings on plain radiography remained normal. All 5 patients were considered to have subsequently developed radiologically confirmed SpA. The mean time period between the study and detection of radiographic sacroiliitis in these patients was nearly 8 years (range 1–18 years). Four patients were already classified as having SpA according to the clinical criteria, while 1 patient was HLA–B27 negative and was initially not classified as having SpA (Figure 1).
In this population-based cohort of patients with ≥3 months of back pain and onset at age <45 years, we found that 8.4% fulfilled the ASAS classification criteria for SpA. Almost 3 times as many patients fulfilled the clinical arm of ASAS-defined SpA as compared to the radiographic arm, indicating that the high HLA–B27 population prevalence (16%) had a significant impact on axial SpA frequency. We further show that at least 20% of patients in the clinical arm subsequently fulfilled the radiologic SpA criteria.
The presence of definite inflammatory SI changes on radiographs was 2.8%, which approximates the findings by O'Shea et al of definite AS changes in 3.8% of patients in a similar but smaller cohort, and our data also confirm the male predilection in radiographically confirmed AS (12). Our data therefore support the validity of the radiograph-based arm of SpA among patients with CBP, which is in line with its applicability in established SpA cohorts (13).
Based on the cohort size data of the original survey, we estimate the total prevalence of undiagnosed SpA in this survey to be 0.13%. This estimate for clinical SpA alone reached 0.1% and reached 0.03% for radiographic SpA alone. Taken together with our earlier data on the prevalence of established radiographic SpA of 0.4% in this region, this would bring the total SpA prevalence to 0.53%. This falls within the range of SpA prevalence rates described from various parts of the world (14, 15).
MRI has increasingly been accepted as a diagnostic tool in the evaluation of patients with IBP (16). However, MRI changes are not fully equivalent to sacroiliitis by radiographs and also do not necessarily indicate subsequent development of permanent damage with erosions and sclerosis in all patients. Bennett et al found that only one-third of patients demonstrating sacroiliitis by MRI at baseline after 8 years developed sacroiliitis by radiographs, indicating that bone marrow edema may resolve (6, 17). Also, access to MRI is not universally available to patients due to costs or availability issues. This indicates the need for an alternative way of SpA classification that circumvents the restrictions imposed by radiography. Previous SpA criteria, such as the Amor and European Spondyloarthropathy Study Group (ESSG) criteria, already recognized this need for a broader definition of SpA to allow earlier diagnosis. Whereas the ESSG criteria were anchored around inflammatory spinal pain and 1 additional SpA feature, the Amor criteria included weighting of 10 different SpA features, with a final score of 5 indicating probable SpA and a score of 6 indicating definite SpA (18, 19). Both classification sets allow, but do not require, radiographic evidence of sacroiliitis to be present and both perform as well as the newer ASAS classification criteria in discriminating patients with established disease (13). However, the ESSG and Amor criteria sets have not been widely implemented, possibly due to the lack of an effective intervention at the time. In the clinical arm of the ASAS classification criteria, HLA–B27 was selected as the anchoring feature because of its sensitivity, specificity, and face validity for SpA, and because the initial use of IBP as the anchoring feature was found to lead to poor specificity of the data set (3). In axial SpA, the lack of a gold standard for early diagnosis and the increasing availability of effective antiinflammatory therapy have resulted in the development of algorithms that “facilitate an earlier diagnosis.” These algorithms aim to reduce the shaded area between classification and diagnosis. Even though algorithms include many of the established classification criteria, they weigh these criteria differently according to existing knowledge/insights on their sensitivity/specificity (20, 21). We therefore did investigate how well the separate features of the clinical ASAS classification criteria perform in distinguishing patients with axial SpA from patients with CBP. The high diagnostic OR for HLA–B27 (Table 3) supports its use as an anchor criterion. The rationale for the use of a diagnostic OR is that it provides a single indicator of test performance that is only dependent on the control group and independent of prevalence, while its format is familiar to medical practitioners and higher diagnostic ORs are indicative of better test performance (22). We also show, however, that the diagnostic precision (positive predictive value) of HLA–B27 for SpA is only 48%. The positive predictive value of a diagnostic test course greatly depends on population prevalence, which was nearly 16% in our cohort, whereas the ASAS study cohort contained undisclosed numbers of patients from regions with a low (1–5%) to moderate (6–9%) prevalence of HLA–B27.
On the one hand, early diagnosis is paramount to improve the long-term functional outcome in SpA patients (23), and given the risk for work disability in the prototypical axial SpA (16, 24) as well as the relationship between ongoing disease activity and cardiovascular events in AS (25, 26), it is obvious that any tool that will shorten the delay to treatment will be beneficial. Although not specifically addressed in the latest recommendations for biologic treatment for axial SpA (27), further research will be needed before clinical SpA criteria can be translated into clinical practice, especially with regard to biologic treatment. The feasibility and long-term outcome of biologic treatment as compared with standard therapy for clinical axial SpA have yet to be documented in a satisfactory manner, making it uncertain whether patients with a clinical diagnosis of axial SpA will eventually have the intended better outcome (28). This is especially important in view of the possible consequences for health care resources in general.
Limitations of this study should be mentioned. Our data are based on a cohort of patients with self-reported back pain. While we excluded patients with non–ASAS-defined CPB, we cannot be absolutely sure that our final cohort is fully representative of standard patients referred to a rheumatologist for SpA evaluation. Our data set lacks MRI findings, which have increasingly been accepted as a diagnostic tool to demonstrate ongoing sacroiliitis, since application of the modified New York criteria incurs an inappropriate diagnostic delay (4, 7). It is therefore unclear how our results would be affected had MRI data been available. The subsequent development of radiographically confirmed SpA in 5 patients should be considered a minimum number because they are based on hospital registration data, which have not necessarily captured all subsequent cases.
We show a considerable increase in SpA prevalence when clinical ASAS criteria for SpA are applied for CBP patients in regions with a high prevalence of HLA–B27. Although no definite role for biologic treatment has been defined in patients with clinical SpA, these data indicate that such a strategy could have significant ramifications for health care delivery.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Nossent had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Bakland, Alsing, Singh, Nossent.
Acquisition of data. Alsing, Singh, Nossent.
Analysis and interpretation of data. Bakland, Nossent.
The authors wish to thank M. Andersen and D. Nossent for their assistance with data compilation.