Dr. van der Heijde has received consulting fees, speaking fees, and/or honoraria from Abbott, Amgen, Centocor, Wyeth, Pfizer, UCB, Schering-Plough, Eli Lilly, Bristol-Myers Squibb, Novartis, Roche, and MSD (less than $10,000 each) and is the director of Imaging Rheumatology.
The term nonradiographic axial spondyloarthritis (SpA) has recently been introduced to identify patients with axial SpA earlier, before structural changes in the sacroiliac joints have been detected. This was an important step forward for earlier diagnosis and for the conduct of treatment trials in these patients. With the new definition of nonradiographic axial SpA as part of the axial SpA group, several questions have been raised regarding this subgroup, which will be discussed in this review.
The historical development of axial spondyloarthritis (SpA)
The term ankylosing spondylitis (AS) was created around 1900, a time when a diagnosis could only be made clinically, after a patient had already developed advanced ankylosis of the spine. With the increased use of radiography, it became clear during the 1930s that the disease normally starts in the sacroiliac (SI) joints (1). Based on this feature, radiographic sacroiliitis has become an important part of the diagnosis of AS and is an essential criterion of the modified New York criteria, which were published in 1984 (2). Yet, the term AS was already inadequate because these criteria could be fulfilled based only on the presence of radiographic sacroiliitis, with or without ankylosis of the spine. Moreover, radiographs become positive for sacroiliitis only at a rather late stage, since they detect only structural damage as a consequence of inflammation and not the inflammation itself, though the name itself suggests otherwise (3). Thus, cases of early inflammatory disease were not covered by these criteria.
This stage of the disease—before the occurrence of radiographically visible damage of the bone—has recently been termed nonradiographic axial SpA (3, 4). The existence of such cases, however, has been known for a long time (although by different nomenclature), certainly since the publication of the European Spondylarthropathy Study Group (ESSG) criteria in 1991 (5) and the Amor criteria in 1990 (6). Numerous investigations in patients with nonradiographic axial SpA have been performed over the last 10–20 years, mostly using the ESSG criteria (or stricter definitions). While the term “undifferentiated” SpA was frequently used to describe the disease (5), it could easily be misinterpreted as representing “not well defined” SpA.
The new Assessment of SpondyloArthritis international Society (ASAS) criteria were subsequently developed as classification criteria for axial SpA (Figure 1), covering both radiographic axial SpA (radiographic sacroiliitis according to the modified New York criteria) and nonradiographic axial SpA (4). Expert opinion on SpA versus non-SpA was used as the gold standard for these criteria. The main purpose for the development of these new criteria was to include magnetic resonance imaging (MRI) assessments and to improve the sensitivity and specificity of the criteria as compared to the available criteria, such as the ESSG or Amor criteria. Indeed, the new ASAS criteria had a sensitivity of 82.9% and a specificity of 84%, as compared to 70.7% and 63.5%, respectively, for the ESSG criteria and 69.4% and 78.4% for the Amor criteria. Moreover, these criteria allow a clearer distinction between axial and peripheral SpA as compared to the ESSG and Amor criteria. Having groups that are more homogeneous would also have advantages for future etiologic and genetic studies. These criteria do not define/describe a new group of SpA patients, but rather, they improve the classification of these patients.
What are the frequencies of patients with nonradiographic axial SpA versus radiographic axial SpA (or, AS according to the modified New York criteria)?
Most of the available data are from studies performed in patients with suspected axial SpA, but without a clear diagnosis, who were referred to rheumatologists. Axial SpA was then diagnosed for the first time. The characteristics of these studies and the frequencies of nonradiographic axial SpA to AS identified are summarized in Table 1. The data indicate that the proportion of nonradiographic axial SpA patients among those with newly diagnosed axial SpA can be expected to be between 23% and 80%, depending on the symptom duration, selection criteria, and other parameters, such as availability and interpretation of MRIs. Interestingly, some of the studies showed that radiographic sacroiliitis may already be present in 20–30% of patients after only 2–3 years of symptom duration (7–10).
Table 1. Ratio of nonradiographic axial SpA to AS in patients with axial SpA diagnosed by a rheumatologist after referral*
No. of axial SpA patients
Reason for referral
% with nonradiographic axial SpA/AS
SpA = spondyloarthritis; AS = ankylosing spondylitis; ASAS = Assessment of SpondyloArthritis international Society; SPACE = Spondyloarthritis Caught Early; DESIR = Devenir des Spondyloarthropathies Indifferérenciées Récentes; IBP = inflammatory back pain; MASTER = Multicenter Ankylosing Spondylitis Survey Trial to Evaluate and Compare Referral Parameters in Early SpA; RADAR = Recognising and Diagnosing Ankylosing Spondylitis Reliably; ESPAC = Early Spondyloarthritis Clinic.
ASAS classification study
6.1 years (mean)
Chronic back pain, possible SpA
>3 months, <2 years
Chronic back pain
>3 months, <3 years
IBP plus suspected SpA
Berlin referral study
>7 years (mean)
Chronic back pain plus features of SpA
MASTER referral study
>7 years (mean)
Chronic back pain plus features of SpA
RADAR referral study
>7 years (mean)
Chronic back pain plus features of SpA
>3 months, <2 years
Herne referral study
3.7 years (mean)
There are also data from 2 trials of tumor necrosis factor (TNF) blockers that included patients with active axial SpA (11, 12). In the first trial, which consisted of patients with ≤3 years of symptom duration (mean 15.3 months) at study inclusion, the proportion of patients with AS was 12% and the proportion of patients with nonradiographic axial SpA 88% (11). In the other trial, which consisted of patients with active disease and a symptom duration of ≤5 years (mean 2.8 years), the proportion of patients with AS was 51% and the proportion of patients with nonradiographic axial SpA was 49% (12). In a third drug trial, axial SpA patients with a symptom duration of <5 years were included in a multicenter placebo-controlled study of sulfasalazine (13). Only 13% of the total cohort had radiographic sacroiliitis and fulfilled the modified New York criteria, with 87% being classified as having nonradiographic axial SpA.
In a further investigation, 329 patients with axial SpA were investigated for the presence of radiographic sacroiliitis dependent on the duration of symptoms. SpA was diagnosed because all patients had family members with a defined SpA (100% with a positive family history), had complaints of chronic inflammatory back pain (97%), and were positive for HLA–B27 (97%) (14). According to a recently suggested diagnostic approach (3), these patients would have a >90% probability of having axial SpA. When radiographs of the SI joints were obtained, radiographic sacroiliitis (according to the modified New York criteria) was present in 40% of the patients with a symptom duration of <10 years, 70% of those with a symptom duration of 10–20 years, and 86% of those with a symptom duration of >20 years, clearly confirming the time dependency of the occurrence of radiographic sacroiliitis. But, these data also indicate that there seems to be a proportion of patients with axial SpA, ∼10–15%, who probably will never develop radiographic sacroiliitis.
All studies discussed above were focused on patients seen and/or selected by rheumatologists and might not reflect the true ratio of nonradiographic axial SpA to AS at the level of the general population. An earlier investigation studied the prevalence of AS and undifferentiated axial SpA (equivalent to the nonradiographic axial SpA group) among blood donors with and without HLA–B27 and found a nearly similar frequency of the two subgroups (15). Recently, an attempt was undertaken to estimate the US prevalence of inflammatory rheumatic diseases, including spondyloarthritis, based on published studies of smaller, defined populations (16). It was calculated that there was a somehow lower prevalence of undifferentiated SpA (which would include axial undifferentiated SpA [or, nonradiographic axial SpA] and peripheral undifferentiated SpA; these two groups were not distinguished) than AS (prevalence 374 per 100,000 and 520 per 100,000, respectively). Subsequently, a cross-sectional survey was conducted in the US in a representative sample of 5,013 adults, resulting in an estimated prevalence of axial SpA of up to 1% in the US population (17). As a major limitation of that study, no information on SI joint imaging, HLA–B27 status, family history, or previous urogenital or gastrointestinal infections were obtained. Moreover, a ratio between the two subgroups could not be calculated because no radiographs had been performed. But based on better estimates about the US prevalence of AS, it can be postulated that a significant proportion of these patients belong to the nonradiographic axial SpA subgroup.
All of these data from different kinds of studies clearly indicate that patients with nonradiographic axial SpA are a relevant subgroup and that their percentage among the whole group of patients with axial SpA can be expected to be between 20% and 80%.
What is the progression rate from nonradiographic axial SpA to AS?
The data discussed above indicate that the development of radiographic sacroiliitis is time dependent and allow a rough estimation about the speed of this development. But, there are also several studies that address in more detail the question of progression from nonradiographic axial SpA to radiographic axial SpA over time. Sampaio-Barros et al reported a 10% progression rate over 2 years (18) and a 24.3% progression rate over 5–10 years (19). Earlier studies reported progression in 25% of patients over 2–6 years (20) and in 59% of patients over a period of 10 years (21). A more recent investigation from the German Spondyloarthropathy Inception Cohort (GESPIC) addressed this question in patients with nonradiographic axial SpA and a symptom duration of ≤5 years (22). In that study, the rate of radiographic progression over 2 years was 11.6%. Thus, a progression rate of ∼10% over 2 years was reported consistently in all of these studies. C-reactive protein (CRP) positivity was a clear risk factor for radiographic progression in the latter analysis, resulting in a 2-year progression rate of 24% in CRP-positive patients (22).
While nonradiographic axial SpA was defined in these studies clinically and radiographically, there are currently also 3 publications using MRI investigations at baseline. In one of the studies, ∼30% of patients with active inflammation of the SI joints, as shown by MRI at baseline, developed sacroiliitis after 3 years (23). In another followup study of 40 patients with early (symptom duration <2 years) inflammatory back pain, 77% of the patients who developed radiographic sacroiliitis (AS) had positive findings on MRI at baseline, and 62% of those with sacroiliitis on baseline MRI developed radiographic sacroiliitis after a mean followup of 7.7 years (24). In the study from China (25), 87% of patients with positive findings on MRI of the SI joints at baseline developed radiographic sacroiliitis over a followup of 5–10 years, compared to only 30% of patients with negative MRI findings at baseline. Thus, the rate of progression to radiographic sacroiliitis was ∼20% over 2 years in patients with positive MRI findings at baseline in these 3 studies.
Sufficient data are therefore available indicating that the progression rate from nonradiographic to radiographic axial SpA over 2 years is ∼10% and is probably ∼20% in patients with active inflammation on MRI of the SI joints and/or CRP positivity at baseline. Based on these data from both cross-sectional and followup analyses, a certain ratio of nonradiographic axial SpA to AS, depending on specific symptom durations, can be assumed (Figure 2). The exact proportion in relation to symptom duration, however, is unknown.
Is disease activity similar in patients with nonradiographic axial SpA, or does it differ? Does it differ in its response to treatment?
Data from the German GESPIC study of patients with a symptom duration of <5 years (26), from the French Devenir des Spondyloarthropathies Indifferérenciées Récentes (DESIR) cohort of patients with a symptom duration of <3 years (7), and from another observational cohort in which 68% of patients had a mean symptom duration of >5 years (27) indicate that patients have the same level of disease activity, as measured by the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), and the same level of pain, independently of whether they are classified as having nonradiographic axial SpA or AS (Table 2).
Table 2. Comparison of features in patients with nonradiographic axial SpA versus AS patients in 3 noninterventional cohorts*
All AS, <10 years
AS <5 years
Nonradiographic axial SpA <5 years
Nonradiographic axial SpA
Axial SpA, including AS
Data from 3 noninterventional cohorts were analyzed: the German Spondyloarthropathy Inception Cohort (GESPIC) (26), the Herne, Germany, cohort (27), and the Devenir des Spondyloarthropathies Indifferérenciées Récentes (DESIR) cohort (7). In the DESIR study, nonradiographic axial spondyloarthritis (SpA) was not separately reported. AS = ankylosing spondylitis; BASDAI = Bath Ankylosing Spondylitis Disease Activity Index; CRP = C-reactive protein; MRI = magnetic resonance imaging; NA = not assessed; SI = sacroiliac.
There is evidence from the same studies, however, that nonradiographic axial SpA patients have lower CRP levels (26, 27) and less active bony inflammation on MRI (27) as compared to AS patients with a similar duration of symptoms (Table 2). This might be explained by the finding that axial SpA patients with elevated CRP levels and/or positive MRI findings seem to progress faster to the radiographic sacroiliitis stage, and therefore, a higher proportion of AS patients might be found if objective parameters of inflammation, such as CRP or MRI positivity, are present, as discussed above. Both the ASAS recommendations for the treatment of axial SpA with TNF blockers (28) and the European Medicines Agency (EMA) in its recent labeling of adalimumab for the indication of nonradiographic axial SpA require, in addition to a certain level of clinical disease activity, an elevated CRP level and/or MRI score for active bone inflammation to start a TNF blocker in patients with nonradiographic SpA. In these patients, the rates of response to TNF blocker therapy are similar to those in AS patients with the same symptom duration. Such a similar response rate for the two groups could be shown both in direct comparison for etanercept in the Etanercept versus Sulfasalazine in Early Axial Spondyloarthritis (ESTHER) study (12) and in comparison to previous AS trials with adalimumab (AS in the Adalimumab Trial Evaluating Long-term Efficacy and Safety for Ankylosing Spondylitis [ATLAS] trial) versus nonradiographic axial SpA in the study by Haibel et al or the Ability-1 study (29–31), or with infliximab in the study of early axial SpA by Barkham et al (11) versus infliximab in AS (32). In support of this, the ASAS criteria for 40% improvement in disease activity (ASAS40) response rate did not differ between the adalimumab and the placebo groups in patients with nonradiographic axial SpA who were negative for both CRP and active inflammation on MRI (23% versus 20%) in the Ability-1 study (31). Thus, short symptom duration and high disease activity matter the most with regard to the response to TNF blockers, while the absence or presence of structural damage of the SI joints as shown on radiographs does not seem to be a differentiating factor for treatment responses.
Furthermore, in a prospective cohort of patients with axial SpA (741 patients with AS and 162 with nonradiographic axial SpA) from Switzerland, where fulfillment of the modified New York criteria for AS is not mandatory for TNF blocker therapy, it was shown that TNF blockers were used to treat similar percentages of patients with AS and with nonradiographic axial SpA (74% and 71%, respectively) in groups matched for levels of disease activity (33).
Interestingly, the male-to-female ratio clearly differs between the two subgroups, with a ratio of 2:1 in AS but only 1:1 in nonradiographic axial SpA, which might indicate that women, for unknown reasons, develop structural changes later and/or less frequently than men (Table 2). However, in contrast to the presence of objective signs of inflammation, male sex was not a predictor of a better response to treatment with TNF blockers in nonradiographic axial SpA trials or in AS trials (31). A tendency toward slightly lower rates of HLA–B27 positivity in nonradiographic axial SpA patients compared to AS patients was reported in some investigations, although not in all comparisons (Table 2).
What is the relationship of nonradiographic axial SpA to other defined SpA subgroups, such as reactive arthritis/reactive spondyloarthritis, psoriatic arthritis/spondylitis, or arthritis/spondylitis associated with inflammatory bowel disease (IBD)?
Back pain, often presenting as inflammatory back pain with symptoms at rest and improving with exercise, can occur in up to 40% of patients with acute reactive arthritis (34). These symptoms are mostly transient, however, and do normally not persist for more than 3 months. Reactive arthritis as an initial event in the development of AS, or axial SpA may occur, especially if the patients are HLA–B27 positive (35). But, this is a rather rare event. This small group of patients is normally not listed in the description of clinical characteristics of AS or axial SpA cohorts.
The occurrence of inflammatory back pain symptoms has been reported in as many as 10% of IBD cohorts, but again, this is mostly transient and presents rarely as chronic back pain lasting for more than 3 months (36). Nonetheless, the association of AS and axial SpA with psoriasis and IBD is more frequent than the association with reactive arthritis, with psoriasis occurring in ∼10% and IBD occurring in ∼5% of AS cohorts.
Currently available data indicate that there is no major difference in the prevalence of concomitant IBD and psoriasis between AS patients and nonradiographic axial SpA patients. A similar percentage was found for the whole group of axial SpA patients in the ASAS SpA classification study (with 70% having nonradiographic axial SpA and 30% having AS): psoriasis in 8.4% and IBD in 3.8% (4). In the GESPIC study, AS patients with a symptom duration of ≤10 years and nonradiographic axial SpA patients with a symptom duration ≤5 years were included. Psoriasis and IBD were present in 10.2% and 2.6% of the AS patients, respectively, and in 9.8% and 1.8% of the nonradiographic axial SpA patients, respectively, with no difference in the AS group between patients with symptom durations of <5 years versus 5–10 years (26). Data from the French DESIR cohort also showed the presence of psoriasis in similar percentages of patients with and those without radiographic sacroiliitis (37). Furthermore, treatment guidelines for psoriatic arthritis published by the Group for Research and Assessment of Psoriasis and Psoriatic Arthritis (GRAPPA) (38) and the more recent European League Against Rheumatism (EULAR) recommendations for the management of psoriatic arthritis (39) refer to the ASAS recommendations for the treatment of AS/axial SpA (28) in cases of axial manifestations, indicating that a subgroup of patients with psoriatic spondylitis is currently not well defined and has not been investigated in treatment trials separately from AS/axial SpA. In a recent publication (40), it was also shown that AS patients with and those without concurrent psoriasis respond similarly well to treatment with TNF blockers.
Therefore, patients with nonradiographic axial SpA in association with reactive arthritis, psoriasis, or IBD should not be viewed differently from AS patients with the same associated conditions. The common factor for these patients is the axial manifestations, and there is currently no evidence in favor of splitting the axial SpA patients into other subgroups.
Definition of nonradiographic axial SpA
According to the recently developed ASAS classification criteria for axial SpA, patients can fulfill either the imaging arm (radiographic sacroiliitis or active inflammation of the SI joints on MRI plus one other characteristic feature of SpA) or the clinical arm (positive for HLA–B27 plus 2 additional characteristic features of SpA) and always on the background of having chronic back pain for ≥3 months with an age at onset of <45 years (4). The overall sensitivity and specificity of the ASAS criteria were shown to be 82.9% and 84.4%, respectively, using the expert's opinion of axial SpA (the opinion of the rheumatologist with specific expertise in SpA) as the external standard, which are quite good. But an often-raised question concerns the sensitivity and specificity of the imaging arm as well as of the clinical arm in nonradiographic axial SpA.
The imaging arm, including radiography and MRI, had a high specificity of 97.3%, but the sensitivity was clearly reduced to only 66.2%. Positive findings on MRI of the SI joints is a mandatory requirement for the imaging arm in nonradiographic axial SpA, and therefore, a definition of a “positive MRI” is crucial. The published ASAS definition (41) is based on agreement among a group of rheumatologists and radiologists with expertise in SpA and MRI assessments. The proposed definition was subsequently discussed and voted on by attendees at the annual ASAS assembly in 2008. After this, the ASAS definition was tested in independent cohorts. In the largest investigation performed so far, 187 subjects were analyzed, including patients with AS, patients with suspected nonradiographic axial SpA, and control groups (42). A specificity of 88% and a sensitivity of 67% were found in patients with nonradiographic axial SpA, as compared to the various controls, for the ASAS definition. When assessment of structural changes such as erosions was added, the sensitivity increased to 81%, but the specificity remained the same. Interestingly, the specificity was as high as 97% if the judgment about positive or negative MRI findings typical of SpA was based on a global assessment by experienced readers instead of using a predefined cutoff; however, the sensitivity dropped to 51%. In a smaller study, a similar specificity of 89% was reported for the ASAS definition (43). When stricter criteria for subchondral bone marrow edema were applied, the specificity did not increase (still 89%), but the sensitivity dropped from 89% to 69%. Thus, using the ASAS criteria for positive MRI findings, a specificity of ∼90% can be expected; sensitivity might be improved by adding parameters for structural damage, such as erosions.
Relying only on the imaging arm, however, would miss about one-third of patients with axial SpA because of its low sensitivity. In a recent analysis of the axial SpA classification study (4), the specificity and sensitivity of the clinical arm alone were determined to be 83.3% and 56.6%, respectively (Rudwaleit M, van der Heijde D, Sieper J: unpublished observations), indicating that the specificity of the clinical arm alone is also good, but again, with low sensitivity, as for the imaging arm alone.
There are several reasons why axial SpA patients could fulfill only the criteria for the clinical arm without also fulfilling the imaging criteria. First, MRI might not be readily available and/or might be too expensive in some areas of the world. Indeed, MRI appears normally at the end of proposed diagnostic algorithms for axial SpA, mostly because of this (3, 44, 45). Second, the ASAS classification criteria include only radiographic or MRI changes of the SI joints, but not other sites of the axial skeleton. Depending on the duration of symptoms, inflammation of the spine as shown on MRI is present in 12–70% axial SpA patients (11, 12, 46), and this might also occur in the absence of inflammation of the SI joints (4, 12, 47). Bony inflammation can also occur at any enthesitic site of the pelvis, independently and without concurrent sacroiliitis. Moreover, structural changes of the SI joints may already be present, even in the absence of current inflammation. Currently, these structural changes of the SI joints seen on MRI are not part of the definition of positive findings on MRI. This might change in the (near) future, as suggested recently (48). Finally, when MRI evidence of inflammation of the facet joints (49) or the SI joints (25) was compared with findings on histologic analyses of biopsy samples taken from the same sites, it became evident that MRI might sometimes not be sensitive enough to detect inflammation. Similarly, there is a rather poor correlation between MRI evidence of inflammation and clinical evidence of disease activity (50–52). Because of fluctuations in inflammation or because of an insufficiently high sensitivity, MRI of the SI joints can show negative results at one time point but positive later on (53). MRI-negative patients, who fulfill only the clinical criteria, might progress more slowly or less frequently to radiographic sacroiliitis (43), as was also discussed above (23–25).
We would like to stress that the data presented herein for the ASAS criteria for axial SpA, including their single arms and the definition of positive findings in the SI joints on MRI, were developed as classification criteria and not as diagnostic criteria (4). A diagnostic approach uses the same parameters as those used for classification, but in a different way (3). A diagnostic algorithm for axial SpA that takes into account the pretest probability, the presence of not only positive findings, but also negative findings, and the number of parameters that are positive (3) has recently been proposed by the ASAS (45).
Nonradiographic axial SpA is clinically a relevant subgroup of axial SpA. An increasing number of studies are now available that allow some estimation of the relative proportion of these patients in relation to AS as well as the progression rate from nonradiographic axial SpA to AS. Similarities and differences between the two subgroups have been described. While they are similar in their clinical presentation, level of clinical disease activity, and response to TNF blockade in patients with the same disease activity levels, they differ in their extent of structural damage (by definition), level of objective inflammation, as shown by the CRP values and MRI findings, and sex distribution. Nonetheless, because most of the data presented here are based on cross-sectional investigations or short-term followup investigations future research should focus on further evaluation of the current classification criteria and on collecting more data on the long-term disease course.
Drs. Sieper and van der Heijde drafted the article, revised it critically for important intellectual content, and approved the final version to be published.