New axSpA Criteria: Who gets left out? Save
Transcription
Hello everyone. This is Nelli Zadi from Beirut, Lebanon, reporting for RheumNow from EULAR in London. I will talk about the new expert classification criteria. You've been hearing about them a lot. What is new in this congress is their validation in two cohorts.
So I will start by just setting up the idea. So today Martin Rudwaleit presented the evolution of axSpA criteria and reminded us how we ended up here in the first place. So he opened with the context that the 2009 ASAS axSpA criteria were born out of an unmet treatment need. There were two arms — imaging and clinical arm — and at the time the practical reality was that not all countries had access to MRI. This is why they also needed a clinical arm. So these criteria initially performed well: sensitivity 83%, specificity 84%. However, two problems emerged afterwards. One, the low specificity of 84% became a real clinical issue and the criteria were frequently misused as diagnostic criteria. So clinicians found it easy just to tick boxes rather than to first establish a clinical diagnosis and exclude mimics.
Enter the CLASSIC study. So the CLASSIC study was a landmark joint initiative of ASAS and SPARTAN, recruiting from 61 centers across 27 countries, 50/50 split between North America handled by SPARTAN and the rest of the world handled by ASAS, to address these issues. So they designed — they used the same CRF, the same entry criteria, which is undiagnosed chronic low back pain for more than 3 months, onset before 45 years, and crucially a new MRI approach based on both active and structural lesions and also assessed by central readers. So this was the different thing.
So they tested the 2009 criteria first in over a thousand patients worldwide and they missed the prespecified targets, which were 75% sensitivity and 90% specificity. So when they missed these targets they decided to revise the criteria, and the revised criteria was presented at this congress by Walter Maksymowych, OP0236. He explained the LASSO regression and the multivariable logistic regression that were used — the two methods used to identify the variables most independently associated with a stage-five axSpA diagnosis across over 1,000 patients — and the top finding was that the MRI of the sacroiliac joint by global assessment, active and structural, has the strongest independent association, followed by radiographic sacroiliitis. Other clinical variables selected were HLA-B27, inflammatory back pain, IBD, acute anterior uveitis, heel enthesitis, and dactylitis. An expert consensus then added psoriasis and replaced dactylitis with elevated CRP.
So there were two proposals. There was a vote — the winning criteria, 158 compared to 151 votes, used a weighted point system that was presented at the congress, and you can see it has been published widely — and more than 11 points was the cutoff. So in this validation dataset, using the local reader, sensitivity and specificity were 79% and 90% respectively. The targets were met.
So the key message was — what's new is the central role of imaging and the focused weighted clinical variables. The future challenges remain in MRI interpretation and over-interpretation of MRI, and we really need ASAS-recommended MRI sequences in daily practice.
Now the new thing at this congress is that it was tested in two cohorts. It was tested in the SPACE cohort and in the RABBIT cohort. The SPACE cohort was presented by Jabon Aal, OP0239. So the SPACE cohort provides a critical independent test in early disease. So these are patients with back pain of less than 2 years duration. So this is the hardest test for any classification criteria. They analyzed 669 patients. Mean age was 30. 39% were males. 63% had an axSpA rheumatology diagnosis at two years.
Now the bottom line is that the new 2025 criteria were highly specific — more than 94% — regardless of local or central reader. Sensitivity, however, is meaningfully lower than the 2009 criteria: 60% by local reading versus 83% in 2009, and substantially lower by central reading, like 35 to 52% versus 70 to 78%. The clinical arm is almost emptied out — 83 to 93%, 84 to 93% fewer patients classified via clinical arm only compared to 2009 — and adding the structural MRI lesions to the active lesions gained only a modest plus 1.4 to 5.9% sensitivity with no specificity lost.
So here the trade-off is clear. The new criteria sacrifice some sensitivity — particularly by no longer capturing patients who would have qualified on clinical features alone — in exchange for a major gain in specificity. Interestingly, about 27% of axSpA patients classified by the 2009 criteria are missed by the new one. The reassuring finding is that high specificity is achievable even with local reading, but this is a SPACE cohort and they don't need central readers. But I think the context is very specific to the SPACE cohort.
And the second cohort was a RABBIT registry, presented by
at poster 0190, this is a German registry which brings a very large real-world perspective. 2,236 rheumatologist-diagnosed axSpA patients. 73% fulfilled the 2009 criteria, 67% fulfilled the 2025 criteria, 65% fulfilled both, and 24% fulfilled neither. Again, the imaging arm remains a workhorse — the majority of patients qualify through it in both criteria sets: 68% in 2009, 64% in 2025. Again, here the clinical arm collapses dramatically under the new criteria, from 58% to just 3% in 2025.
So, a notable finding here: 20% of patients failed the entry criteria entirely — no chronic back pain, or onset more than 3 months, or onset over 45 years — which reminds us that one in five real-world axSpA patients may not be reachable by any set of criteria.
So the takeaway: the 2025 ASAS criteria represent a deliberate pivot — specificity over sensitivity — with the imaging arm now firmly in the center. The clinical arm seems no longer to be a route for most patients. This is a paradigm shift for classification. It has real implications for clinical trials, for registries, for access to biologics in certain health systems.
So the criteria performed as expected in the development cohort, whether in early disease, in SPACE, or real-world settings in RABBIT — the trade-offs are really visible and quantifiable. What we need now is better MRI standardization, and probably also a parallel effort to think about how early or seronegative patients without clear imaging can be diagnosed, managed, and classified outside this classification framework. Thank you.
So I will start by just setting up the idea. So today Martin Rudwaleit presented the evolution of axSpA criteria and reminded us how we ended up here in the first place. So he opened with the context that the 2009 ASAS axSpA criteria were born out of an unmet treatment need. There were two arms — imaging and clinical arm — and at the time the practical reality was that not all countries had access to MRI. This is why they also needed a clinical arm. So these criteria initially performed well: sensitivity 83%, specificity 84%. However, two problems emerged afterwards. One, the low specificity of 84% became a real clinical issue and the criteria were frequently misused as diagnostic criteria. So clinicians found it easy just to tick boxes rather than to first establish a clinical diagnosis and exclude mimics.
Enter the CLASSIC study. So the CLASSIC study was a landmark joint initiative of ASAS and SPARTAN, recruiting from 61 centers across 27 countries, 50/50 split between North America handled by SPARTAN and the rest of the world handled by ASAS, to address these issues. So they designed — they used the same CRF, the same entry criteria, which is undiagnosed chronic low back pain for more than 3 months, onset before 45 years, and crucially a new MRI approach based on both active and structural lesions and also assessed by central readers. So this was the different thing.
So they tested the 2009 criteria first in over a thousand patients worldwide and they missed the prespecified targets, which were 75% sensitivity and 90% specificity. So when they missed these targets they decided to revise the criteria, and the revised criteria was presented at this congress by Walter Maksymowych, OP0236. He explained the LASSO regression and the multivariable logistic regression that were used — the two methods used to identify the variables most independently associated with a stage-five axSpA diagnosis across over 1,000 patients — and the top finding was that the MRI of the sacroiliac joint by global assessment, active and structural, has the strongest independent association, followed by radiographic sacroiliitis. Other clinical variables selected were HLA-B27, inflammatory back pain, IBD, acute anterior uveitis, heel enthesitis, and dactylitis. An expert consensus then added psoriasis and replaced dactylitis with elevated CRP.
So there were two proposals. There was a vote — the winning criteria, 158 compared to 151 votes, used a weighted point system that was presented at the congress, and you can see it has been published widely — and more than 11 points was the cutoff. So in this validation dataset, using the local reader, sensitivity and specificity were 79% and 90% respectively. The targets were met.
So the key message was — what's new is the central role of imaging and the focused weighted clinical variables. The future challenges remain in MRI interpretation and over-interpretation of MRI, and we really need ASAS-recommended MRI sequences in daily practice.
Now the new thing at this congress is that it was tested in two cohorts. It was tested in the SPACE cohort and in the RABBIT cohort. The SPACE cohort was presented by Jabon Aal, OP0239. So the SPACE cohort provides a critical independent test in early disease. So these are patients with back pain of less than 2 years duration. So this is the hardest test for any classification criteria. They analyzed 669 patients. Mean age was 30. 39% were males. 63% had an axSpA rheumatology diagnosis at two years.
Now the bottom line is that the new 2025 criteria were highly specific — more than 94% — regardless of local or central reader. Sensitivity, however, is meaningfully lower than the 2009 criteria: 60% by local reading versus 83% in 2009, and substantially lower by central reading, like 35 to 52% versus 70 to 78%. The clinical arm is almost emptied out — 83 to 93%, 84 to 93% fewer patients classified via clinical arm only compared to 2009 — and adding the structural MRI lesions to the active lesions gained only a modest plus 1.4 to 5.9% sensitivity with no specificity lost.
So here the trade-off is clear. The new criteria sacrifice some sensitivity — particularly by no longer capturing patients who would have qualified on clinical features alone — in exchange for a major gain in specificity. Interestingly, about 27% of axSpA patients classified by the 2009 criteria are missed by the new one. The reassuring finding is that high specificity is achievable even with local reading, but this is a SPACE cohort and they don't need central readers. But I think the context is very specific to the SPACE cohort.
And the second cohort was a RABBIT registry, presented by
at poster 0190, this is a German registry which brings a very large real-world perspective. 2,236 rheumatologist-diagnosed axSpA patients. 73% fulfilled the 2009 criteria, 67% fulfilled the 2025 criteria, 65% fulfilled both, and 24% fulfilled neither. Again, the imaging arm remains a workhorse — the majority of patients qualify through it in both criteria sets: 68% in 2009, 64% in 2025. Again, here the clinical arm collapses dramatically under the new criteria, from 58% to just 3% in 2025.
So, a notable finding here: 20% of patients failed the entry criteria entirely — no chronic back pain, or onset more than 3 months, or onset over 45 years — which reminds us that one in five real-world axSpA patients may not be reachable by any set of criteria.
So the takeaway: the 2025 ASAS criteria represent a deliberate pivot — specificity over sensitivity — with the imaging arm now firmly in the center. The clinical arm seems no longer to be a route for most patients. This is a paradigm shift for classification. It has real implications for clinical trials, for registries, for access to biologics in certain health systems.
So the criteria performed as expected in the development cohort, whether in early disease, in SPACE, or real-world settings in RABBIT — the trade-offs are really visible and quantifiable. What we need now is better MRI standardization, and probably also a parallel effort to think about how early or seronegative patients without clear imaging can be diagnosed, managed, and classified outside this classification framework. Thank you.
If you are a health practitioner, you may Login/Register to comment.
Due to the nature of these comment forums, only health practitioners are allowed to comment at this time.