Disease classification criteria aim to define cases of high clinical and aetiopathogenic similarity to facilitate clinical and basic research which otherwise would be adversely affected by disease heterogeneity. The more that is understood about the pathogenesis of diseases, generally the greater our ability to subdivide them into more homogeneous subsets, thereby enabling better prediction of natural history, facilitating research into aetiopathogenesis, and identifying effective treatments. In rheumatology, recent examples of major advances through tighter disease classification include the division of rheumatoid arthritis into anti-citrullinated protein antibody positive and negative disease, and the division of vasculitis according to anti-neutrophil cytoplasmic antibody status. These improvements in classification have come about through research demonstrating heterogeneity (including in genetic and environmental aetiopathogenesis, immunology, natural history and treatment response) in cases meeting previous classification criteria, and have benefited both research and clinical practice.

The modified New York (mNY) criteria for ankylosing spondylitis (AS) have long been recognised to have low sensitivity to establish a diagnosis of AS in early disease. There have been many attempts to produce improved criteria with greater sensitivity in early spondyloarthritis, including most recently the development of the Assessment of SpondyloArthritis International Society (ASAS) axial spondyloarthritis (axSpA) criteria.1 ,2 While these criteria have been developed with honourable intent, we argue here that they have been developed without sufficient reference to what we know of the natural history of axial SpA and its underlying aetiopathogenic mechanisms. As a consequence, while they capture a higher proportion of patients with axial SpA than previous criteria, they do so at a cost of substantially increasing the heterogeneity of the resulting disease group and reducing the utility of these criteria for both clinical and research applications.

The diagnosis of AS has been defined historically by the Rome, New York, and then most recently by the mNY criteria.3 ,4 The salient features of these sets of criteria have been the presence of sacroiliac joint plain radiographic change and symptoms and signs of axial arthritis. The mNY criteria are specific, in that a patient who can be classified as AS by these criteria really does have the disease entity which rheumatologists recognise as AS.4 However, the requirement for radiographic sacroiliitis clearly reduces the sensitivity of the mNY criteria, particularly in early disease, contributing at least in part to the often considerable diagnostic delay in AS.5 The introduction of MRI scanning in SpA has confirmed the long held belief that there is a very large group of patients with inflammatory axial SpA who do not yet have radiographic changes. These patients are currently largely officially excluded from public funded programmes of tumour necrosis factor (TNF) antagonist treatment, although there is good evidence that carefully selected cases do respond to such treatment.6

It has been proposed that SpA patients with predominately axial symptoms should be considered as having the same disease entity as AS, and that all patients with axSpA are essentially early cases of AS.7 To further advance this proposal, a new disease concept for SpA patients was created which included both those with, and those without radiographic sacroiliitis.7 The ASAS group then developed criteria that included non-radiographic or pre-radiographic axSpA.1 ,2 There is, however, a growing body of evidence which suggests that the two entities of axSpA and AS are different, and that while a currently unknown proportion of patients with axSpA do indeed have early AS, this is far from universal.

A compelling paper described the evidence that axSpA and AS are in fact the same disease entity,7 and there has been considerable debate within the AS research community as to whether the term AS should be replaced with axSpA. We argue that the term AS should be retained, because axSpA cases have far greater clinical heterogeneity than AS, and have a broader aetiopathogenesis; furthermore, the natural history of axSpA has not yet been reliably established—it clearly varies more between cases than does AS.

Why is this important to make this distinction? For both clinical practice and biological research into AS and spondyloarthritis, it is important that we accurately classify disease. The risk of misdiagnosis has been rightly highlighted when using classification criteria to make diagnoses.7 Despite this, classification criteria are commonly used in clinical practice to make diagnoses and can lead to misdiagnosis. The axSpA criteria have already being presented as diagnostic criteria in rheumatology practice.8 ,9 Treatments, especially with immunosuppressants, are not without risk, can be extremely expensive, and their risk/benefit ratio is affected by knowledge of prognosis. In addition, our ability to advance our knowledge of the biology of AS and axSpA depends critically upon assembling homogeneous groups of patients to study.

Although AS and axSpA are clearly overlapping entities, there are also many differences between them. Considering natural history, few longitudinal studies have assessed the risk of ultimately developing AS in cases of axSpA. Mau et al followed 88 patients with suspected AS not meeting mNY criteria for 10 years.10 After 10 years, 54 patients remained in the study, of whom 22 (41%) had not reached the New York criteria for AS; indeed 12 (22%) had AS excluded and an alternative diagnosis made, such as psoriatic arthritis, degenerative disc disease or fibromyalgia.10 The Leeds Inflammatory Back Pain (IBP) study followed 29 patients with symptoms suggestive of axSpA and positive MRI scans, meeting the ASAS axSpA criteria; a mean follow-up of 7.7 years has been published.11 Eight patients fulfilled the mNY criteria at baseline, but of the remaining 21, only three had developed mNY AS at follow-up. Including all patients, at follow-up 11 had mNY AS, two had psoriatic spondyloarthritis (not further defined), five had reactive arthritis (ReA) and 11 had undifferentiated SpA. The study found that the imaging arm of the axSpA criteria has 100% sensitivity but low (22%) specificity for mNY AS, and the HLA-B27 arm had 67% sensitivity and 56% specificity for mNY AS. Ten of the 21 patients showed radiographic progression of their sacroiliitis, but insufficient to fulfil the mNY criteria after 8 years. Thus, while we do not yet have good data about the natural history of axSpA, it is clear that the diversity of outcomes is far greater in cohorts with axSpA than with mNY AS.

A number of studies demonstrate that AS and axSpA differ in their genetics. First, examining the German Spondyloarthritis Inception Cohort (GESPIC),12 the rates of HLA-B27 carriage differ between the non-radiographic axSpA and AS (72.6% vs 84.3%, respectively). In the ABILITY-1 trial of adalimumab conducted on patients who fulfilled the ASAS axSpA criteria, only 58% of the MRI imaging arm were HLA-B27 positive, whereas in most AS cohorts, >80% of patients are HLA-B27 positive.6 In another trial of adalimumab in axSpA, patients were required to have two out of three features (IBP, HLA-B27 or acute inflammation on MRI). In this trial the HLA-B27 rate was only 67%.13 AS is a highly heritable disease; these data indicate that for the major gene causing AS (HLA-B27), axSpA has a significantly weaker association, suggesting greater heterogeneity of genetic risk underlying axSpA. HLA-B27 carriage itself has been shown not to be associated with the severity of radiographic disease in AS,14 and thus the lower prevalence of HLA-B27 in axSpA cohorts is not simply due to ascertainment for cases prone to osteoproliferative disease.

Gender ratios have also been noted to be different between axSpA and AS. Data from the GESPIC cohort show that gender ratios differ significantly, with 34–43% men in the axSpA group and 65% in the AS group.12 ,15 Braun et al recently reported that in a consecutive case series, only 31.8% of those with non-radiographic axSpA were male, compared 76.8% of those with AS.16 As well as differing in their genetics and demographics, these disease groups also differ in their response to treatment. A French study showed that therapeutic response and partial remission rates to TNF inhibitors were significantly different between those with positive imaging and those without.17 Those with positive imaging also have an increased chance of progressing to mNY AS.18 In the ABILITY-1 trial of adalimumab in axSpA, the response to treatment was much lower than in studies in AS despite similar self-reported levels of disease activity.6

Further data from ReA suggest that axSpA and early AS are different though overlapping entities. ReA is thought to be an aberrant immune response triggered by infection, and a significant proportion of cases settle spontaneously, with chronic SpA developing in only a quarter to a third of patients.19 The clinical features of ReA encompass the spectrum of SpA features and 14–80% are HLA-B27 positive, depending on the cohort, mode of ascertainment and causative organism.19 A high proportion would undoubtedly meet the axSpA criteria,2 but two-thirds to three-quarters then spontaneously resolve, in contrast to AS, where the natural history is more persistent.

In conclusion, there is compelling evidence that axSpA and AS are different though overlapping entities associated with different prognoses, demographics, genetics and responses to treatment. The axSpA concept therefore likely captures those who will progress to AS, but also a significant proportion of individuals with other diseases of varying relatedness to AS. This proposed new disease entity is a timely and welcome initiative to encourage recognition of patients with early disease in order to initiate effective treatment early. However, the heterogeneity of cases is high and the natural history poorly characterised. Long-term studies are needed to determine the prognosis of this new axSpA group, and the proportion that will develop into mNY AS; and the clinical, immunological and genetic subgroups that meet the current axSpA criteria must be better dissected to achieve greater homogeneity.4

Research into the biological and genetic basis of AS should, until the outcome of these long-term trials is known, concentrate on patients who fulfil the mNY criteria.4 There may be value in conducting biological research on the remaining patients, but to maximise the chances of success in our research endeavours it is highly desirable to concentrate on homogeneous groups.

It is salient to consider the comment made by Dr Gene Hunder in reference to rheumatic classification criteria: ‘We must keep in mind that classification criteria and diagnoses are not diseases. They are descriptors that change as new knowledge is acquired’.8 We eagerly await this new knowledge to help further refine the concepts of AS and axSpA for the benefit of all our patients.