Classification Criteria

Introduction

Classification criteria are designed to define for clinical and epidemiological studies a highly disease-specific group of patients. The first criteria for ankylosing spondylitis (AS) were based on the clinical experience of rheumatologists at a meeting held in Rome in Italy in 1961, but since then our understanding of the disease demographics has been changing resulting in subsequent revisions and also new criteria that are all listed in Table 2-1. Thus the Rome criteria were revised at a meeting in New York in the US in 1996 by removal of thoracic pain and uveitis that were deemed to have low specificity or sensitivity, resulting in the New York criteria. 

Incorporation of criteria for chronic inflammatory back pain, as proposed by Calin and colleagues in 1977 (Table 2-2), resulted in modified New York (mNY) criteria, first proposed in 1983 and published a year later. They are the most widely used validated criteria to classify AS, with 98% specificity and 83% sensitivity. According to these criteria, a patient can be classified as having definite AS in the presence of at least one of the clinical features (inflammatory back pain, limitation of mobility of the lumbar spine, or limitation of chest expansion) and the radiologic evidence of definite sacroiliitis.

Diagnostic criteria for AS have also been proposed but they have not been properly validated. Amor Criteria (Table 2-3),  published in 1990, and the European Spondyloarthropathy Study Group (ESSG) criteria published a year later (Table 2-4), were developed encompassing the wider clinical spectrum of spondyloarthritis (SpA) that facilitate earlier disease recognition.

The availability of magnetic resonance imaging ( MRI) with its ability to detect early inflammatory changes in the sacroiliac joint for early recognition of axial spondyloarthritis (axSpA), the advent of new and more effective therapies, and the need to separately identify axial and peripheral forms of SpA were the reasons for the most recently proposed criteria by the Assessment of Spondyloarthritis International Society (ASAS) classification criteria for axSpA in 2009 (Table 2-1) and 2 years later for peripheral SpA.

The ASAS axSpA criteria were developed using a cohort of 649 patients with chronic back pain referred to rheumatologists for suspicion of axSpA. The initial criteria based on roughly 40% of the cohort were subsequently validated by using the remainder (60%) cohort, utilizing “expert” rheumatologists’ opinions. The final criteria for axSpA, the concise form of which is shown in Table 2-5, is based on two sets. One set utilizes the clinical and imaging (by conventional pelvic radiography or by MRI) findings and the other is based on the HLA-B27 status and the clinical findings.

The presence of sacroiliitis (by radiography or by MRI) plus at least one SpA feature (imaging arm) or the presence of HLA-B27 plus at least two SpA features (clinical arm) has 82.9% sensitivity and 84.4% specificity. ASAS has also developed and evaluated the accuracy of the new classification criteria and compared them with the ESSG and the Amor criteria, using the opinion of an expert panel as the reference standard. The ASAS criteria had a sensitivity of 77.8% and a specificity of 82.9%. The modified ESSG criteria had a sensitivity and a specificity of 62.5% and 81.1%, respectively, and the Amor criteria had a sensitivity and a specificity of 39.8% and 97.8%, respectively.

The accuracy of the imaging-arm of the ASAS criteria alone was studied in a case-control study of 48 patients with and without rheumatologist-diagnosed SpA found a sensitivity of only 66% but a specificity of 94%. On the other hand, as mentioned earlier, the mNY criteria set for classification of AS is very highly specific (98% specificity) and very useful clinically if the criteria set is met; but it is not sensitive enough (83% sensitivity) to encompass all patients with AS. The positive predictive value of the confirmation of the initial diagnosis of axSpA after 3 to 5 years of follow-up has been found to be over 90%.

The complex multi-arm selection design of the ASAS classification criteria introduces considerable heterogeneity between patients with radiographic and non-radiographic axial spondyloarthritis (nr-axSpA), and between the imaging and the clinical arm. Application of MRI of the sacroiliac (SI) joints (SIJ) has resulted in a considerably higher prevalence rate of axSpA, along with a higher proportion of females and a lower prevalence figures for HLA-B27 among people classified as axSpA. Data suggest misclassification bias can result in some chronic back pain patients getting falsely labeled as suffering from axSpA. Moreover, the criteria lack, in particular, construct and content validity. Suggestions regarding how to improve the ASAS criteria have been published, and attempts are underway to improve this criteria set. However, it can be stated that these new criteria may enable early recognition of axSpA in patients who present with chronic back pain with onset before age 45, but only after other causes for the patient’s clinical presentation have been excluded.

AxSpA seems to progress to radiographic sacroiliitis relatively more slowly in women than in men. Therefore, among patients classified as nr-axSpA by the ASAS criteria, women comprise >50% of the patients. This confirms the original observation published more than 36 years ago that that women relatively more often present with “spondylitic disease without radiographic evidence of sacroiliitis.” The term “non-radiographic” is currently used to describe this form of axSpA, but it has not been firmly established that nr-axSpA and AS represent one single disease entity because differences between the two entities have been reported regarding gender, HLA-B27 status, burden of inflammation, clinical course and response to anti-TNF treatment. A 35-year follow-up study of a cohort of patients with axSpA and their first-degree relatives revealed considerable heterogeneity of axSpA. One of its major findings was a divergence between AS and nr-axSpA in sex ratios, with a male:female ratio of 2.5:1 for AS, compared to 1:1 for nr-axSpA. Moreover, although data on progression are limited, it appears that not all patients who are diagnosed with nr-axSpA progress to AS, and it may be too early to accept the concept that axSpA is one disease with a spectrum from nr-axSpA to radiographic-axSpA.

A study reported that only a minority (26%) of patients with nr-axSpA progressed to AS when followed for up to 15 years. These authors have therefore stated that “the classification criteria for nr-axSpA identifies many patients who are unlikely to progress to AS,” and they have proposed that nr-axSpA is a prolonged prodromal state that requires longer follow-up to document its evolvement to AS. It has been suggested that nr-axSpA may represent an early stage of AS but may also just be an abortive form of a disease which does cause much pain but which may also never lead to structural changes of the axial skeleton. Moreover, the cut-off between nr-axSpA and AS seems artificial and unreliable, and therefore the term nr-axSpA is much more important for classification than to diagnose patients with axSpA. A latent class and transition analysis conducted in two early axSpA cohorts revealed that there is a considerable overlap between axSpA and peripheral SpA, larger than expected when the ASAS criteria were developed. This analysis, additionally, identified a group of patients representing a grey zone, called “axial SpA at risk.” Of these individuals ≥ 84% fulfilled the ASAS criteria, although they were considered to neither have SpA nor to ever develop it.

Incidentally, the European Medicines Agency (EMA) approved the use of three tumor necrosis factor (TNF) inhibitors (etanercept, adalimumab and certolizumab) for the treatment of patients with nr-axSpA following the initial phase 3 trials conducted in this patient population. However, in United States, the FDA raised several key concerns, such as the uncertainty in the long-term clinical course of this entity and potential misdiagnosis of nr-axSpA in patients with fibromyalgia in the absence of objective signs of inflammation, and did not approve initial applications of adalimumab and certolizumab for the treatment of nr-axSpA. The Food and Drug Administration (FDA) has later approved certolizumab pegol, secukinumab and ixekizumab for the treatment of nr-axSpA, based on clinical trials which addressed and resolved the key issues raised by the FDA.

However, due to the absence of any diagnostic criteria for AS/axSpA, clinicians sometimes inappropriately use the classification criteria for diagnosis. This was unfortunately perpetuated in part by the statement in the abstract of the original paper describing the validation and final selection of the ASAS classification criteria for axSpA that stated that these criteria “may help rheumatologists in clinical practice in diagnosing axSpA in those with chronic back pain.” A recent international survey performed in five countries demonstrated that a substantial majority of rheumatologists are using the classification criteria for diagnostic purpose, while 40% rheumatologists think that the criteria need to be modified. It is of utmost importance to emphasize that the classification criteria and diagnostic criteria differ in several aspects (Table 2-6).

The diagnostic approach is aimed at the estimation of the probability of a suspected disease, whereas the classification approach should be applied to patients with an established diagnosis to define a group, e.g., for clinical and genetic research (Table 2-7). To establish the diagnosis of a disease in clinical practice, we need to exclude other conditions that may explain the patient’s symptoms and such exclusions are not included in the ASAS classification criteria. As clinicians we make decisions about likelihood of a diagnosis that is based on the patient’s clinical history, physical examination, investigations and exclusion of alternative explanations. This decision is not based on whether the patient fulfills the classification criteria. It is hoped that, in near future, advances in our understanding of the biology of axSpA via novel imaging, genetic and biomarker studies will enable the resolution of many current issues in axSpA diagnosis and classification.

Enlarge 

Enlarge 

Enlarge 

Enlarge 

Enlarge 

Enlarge 

Enlarge