Development of the ASDAS

Content validity of the ASDAS was ensured by the first development step, in which a group of experts agreed on a list of domains and individual variables reflecting disease activity that could potentially be included in a new composite index. This was done by a Delphi exercise conducted among ASAS members and resulted in the selection of the following variables: fatigue (BASDAI question 1), axial pain (BASDAI question 2), peripheral pain (BASDAI question 3), duration of morning stiffness (BASDAI question 6), acute-phase reactants (C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR)), patient global assessment of disease activity, swollen joint count (SJC), tender enthesis count, spinal pain, and spinal pain at night .

These items were then tested in the International Study of Starting anti-tumor necrosis factor (TNF) agents in AS (ISSAS) database. In a similar process to the development of the disease activity score (DAS) in rheumatoid arthritis (RA), factor analysis and principle component analysis were used to evaluate the redundancy of items and to reduce the number of variables, and discriminant function analysis and linear regression analysis were used to investigate the optimal weights and combination of variables to be included in the ASDAS . These statistical processes ensured face and construct validity. Initially, four ASDAS formulas were proposed. These were later on reduced by consensus to two formulas, each one of them containing one acute-phase reactant. The preferred formula contains the CRP (ASDAS-CRP) and the alternative formula contains the ESR (ASDAS-ESR) ( Table 2 ). The other variables are shared between the two formulas (but with different weights) and include axial pain (BASDAI question 2), peripheral pain (BASDAI question 3), duration of morning stiffness (BASDAI question 6), and patient global assessment of disease activity.

The discriminative ability of the ASDAS was tested in multiple datasets, including the ISSAS database from where the formulas were developed (external construct: TNF-blocker treatment prescription by the rheumatologist), the Outcome in AS International Study (OASIS) database (external constructs: high and low levels of patient and physician global assessments), and the Norwegian disease-modifying antirheumatic drugs (NOR-DMARD) database (external constructs: high and low levels of physician global assessment, patient-acceptable symptom state, considerable improvement according to the patient, and impact of TNF-blocker vs. DMARD treatment). Overall, these studies have shown that ASDAS performs better than the BASDAI and better than any of the single-item variables, namely individual BASDAI questions, patient and physician global assessments, and acute-phase reactants alone . One of the great advantages of the ASDAS is that it provides a better balance between the patient’s and physician’s perspective, as compared to the BASDAI, which is a fully patient-driven index.

The discriminative ability and sensitivity to change in the ASDAS have also been tested in several clinical trial populations . Results have confirmed the validity of the ASDAS, its highly discriminative ability to detect differences between treatment groups (which has important implications in terms of sample size calculation), and its high sensitivity to change. Since its publication, the ASDAS has been further validated in various populations worldwide . Importantly, ASDAS has been shown to perform as well as or better than existing measures in the entire spectrum of axial SpA, including early and late disease, patients with radiographic and non-radiographic axial SpA , patients with and without peripheral arthritis , patients with and without elevated acute-phase reactants , patients with axial PsA and patients with concomitant fibromyalgia . Interestingly, ASDAS has also been shown to perform well in patients with peripheral SpA , suggesting that it may also be a useful and promising tool to measure and monitor disease activity in patients with peripheral SpA, an application for which it was not initially designed and that requires further scrutiny and validation.

Development of the ASDAS cutoffs and further validation of the ASDAS

Cutoffs are important because they give a meaning to a continuous index. Cutoffs can be defined for the level of disease activity at a single time point (“disease activity states”) or for changes in disease activity after therapeutic interventions (“response criteria”). “Disease activity states” help to decide about the need to change treatment, they can be used as selection criteria for patient participation in research studies, and they can also be used as therapeutic targets (e.g., aiming at remission/inactive disease). “Response criteria” allow measuring the impact of a treatment, namely if the treatment results in clinically relevant improvement. The use of standardized measures across different settings also allows combining results from different studies, for example, for meta-analysis, or to audit results and to define and improve the standards of care.

ASDAS cutoffs were determined in the NOR-DMARD database using receiver operating characteristic (ROC)-analysis. Four disease activity states were defined: inactive disease (a remission-like state), and moderate (although the term “moderate” was chosen, this represents a “low or moderate” disease activity state), high, and very high disease activity. Both the patient and physician global assessments of disease activity at predefined levels (<1, <3, and >6 on a 0–10 scale) were used as external anchors to define the three disease activity cutoffs: 1.3, separating “inactive disease” from “moderate disease activity”; 2.1, separating “moderate disease activity” from “high disease activity”; and 3.5, separating “high disease activity” from “very high disease activity” ( Fig. 1 ). The ASAS partial remission criteria were also used as an additional external anchor for “inactive disease.”

Ankylosing Spondylitis Disease Activity Score (ASDAS) cut-offs for disease activity states (Fig. 1A) and response criteria (Fig. 1B). Cut-offs can be applied to both ASDAS-CRP and ASDAS-ESR; however, the two ASDAS formulas are not interchangeable. A. Disease activity states, B. Response criteria.

Regarding response criteria, the external anchor in ROC-curve analysis was a “global rating of change” after starting treatment, with the health change defined by the patient in five Likert-type categories: “much worse,” “worse,” “unchanged,” “better,” and “much better.” This resulted in the definition of two cutoffs for the magnitude of response: “clinically important improvement” (external construct: patients reporting to be “better” or “much better”), defined as a decrease in ASDAS ≥1.1, and “major improvement” (external construct: patients reporting to be “much better”), defined as a decrease in ASDAS ≥2.0 ( Fig. 1 ).

Validation aspects of the cutoffs were longitudinally tested in the NOR-DMARD database and in a clinical trial population, the AS Study for the Evaluation of Recombinant Infliximab Therapy (ASSERT). ASDAS cutoffs showed excellent psychometric properties, being more discriminative between treatment groups than classical response criteria (BASDAI50, BASDAI 2, ASAS 20, and ASAS 40) ( Table 3 ) . The two currently available remission-like states in axial SpA were also compared, ASDAS inactive disease being more discriminative than ASAS partial remission criteria .

There have been two studies investigating the ASDAS patient-acceptable symptom state (PASS) cutoff. One study suggested cutoff values between 2.5 and 3.0 for ASDAS-CRP and between 2.8 and 3.5 for ASDAS-ESR, depending on the method used to determine the cutoff value . The other study suggested an ASDAS-PASS cutoff ≤2.3, for both ASDAS-CRP and ASDAS-ESR . It is not unexpected that the estimated ASDAS-PASS cutoff is greater that the cutoff between moderate and high disease activity as they represent different concepts and it has been shown that PASS can be associated with substantial levels of pain and disease activity in axial SpA .

The ASDAS and its cutoffs have now been applied in numerous clinical trials and observational studies, including long-term extension studies with TNF blockers, consistently showing good psychometric performance . ASDAS has also received the endorsement of both the ASAS and OMERACT study groups .

Evidence suggests that the ASDAS better reflects the inflammatory disease processes in axial SpA than the BASDAI, namely at the biological (correlation with biomarkers of inflammation, angiogenesis, cartilage, and bone turnover) and magnetic resonance imaging (MRI) level (correlation with MRI inflammation scores) . Furthermore, ASDAS high disease activity (ASDAS ≥2.1) might be a better threshold than the historically used BASDAI elevation cutoff level (BASDAI ≥4) for the selection of patients for treatment with TNF blockers, particularly because it selects a higher number of patients with characteristics predictive of a good response to these therapies .

A longitudinal relationship between disease activity and progression of radiographic damage in AS has recently been shown . This study included patients from the OASIS cohort clinically and radiographically evaluated every 2 years up to a period of 12 years. Several measures of disease activity (ASDAS, BASDAI, and CRP) were significantly associated with an increase in the modified Stoke AS Spine Score (mSASSS), but the ASDAS statistical model was the one that best fitted the data. Interestingly, the effect of ASDAS on mSASSS was higher in males and in patients with less symptom duration. These data add further validity (and predictive value in terms of progression of structural damage) to the ASDAS .

ASDAS feasibility

The ASDAS formula is complex, and it is not possible to mentally calculate the index. However, this is no different from the DAS that has been successfully implemented in RA and, in fact, compared to the DAS, the ASDAS benefits from not requiring a joint count. Rheumatologists are already familiarized with this type of indices and with the strategies that have been put in place to overcome their complexity: the availability of online, desktop, handheld, and smartphone calculators. The ASAS group has developed such tools (available at www.asas-group.org ) for the ASDAS as well as a “quick ASDAS calculation form,” a two-page form that provides the possibility of quickly calculating the ASDAS without the need of an electronic calculator. The availability of these instruments will facilitate the implementation of the ASDAS in clinical practice.

Further insights from clinical practice

In this section, we present two real-life clinical cases and explore the potential advantages of using the ASDAS in clinical practice.

Given the above scenario, rheumatologists would unanimously agree that this patient should be treated with a TNF blocker. However, let us look at the patient’s BASDAI and ASDAS scores:

Interestingly, according to the BASDAI, the patient would not be eligible to start treatment with a TNF blocker, as the BASDAI cutoff of 4.0 is often recommended as the threshold to start this type of treatment . Conversely, according to the ASDAS, the patient has “high disease activity” and would be eligible to start treatment with a TNF blocker as the ASDAS cutoff of 2.1 has been suggested as an alternative to the BASDAI cutoff of 4.0 to start TNF-blocker treatment . Therefore, in the above case scenario, ASDAS better reflects disease activity than BASDAI, from both the patient’s and physician’s perspective. The main reason for the observed discrepancy between ASDAS and BASDAI is that the patient’s complaints are mainly axial (back pain and stiffness), which can be “underrepresented” by BASDAI.

From a clinical perspective, when asked if this patient is in clinical remission/has inactive disease status under TNF-blocker treatment, rheumatologists would unanimously respond “yes” because disease symptoms are completely controlled and the only evidence of disease is the irreversible structural damage that is restricting spinal mobility and function independently of disease activity. However, looking at the two remission-like states that have been defined in axial SpA ( Table 3 ), this patient is only in clinical remission according to the ASDAS remission-like state (ASDAS <1.3, which represents inactive disease) but not according to the ASAS partial remission criteria, which also require a BASFI score <2 (0–10 scale), a goal that is often unrealistic in patients with extensive spinal structural damage. Therefore, for the purpose of defining a remission-like disease activity state, ASDAS is more appropriate than other currently available definitions as the definition of ASDAS inactive disease is independent of physical function.

Importantly, in a recent study reporting the outcomes of TNF-blocker treatment over a period of 2 years, achievement of ASDAS inactive disease or ASDAS major improvement was significantly associated with greater improvements in the 36-Item Short Form Survey (SF-36) physical and mental component scores as well as in work productivity compared to patients who did not meet these treatment targets . These data show that ASDAS response translates into improvements in health-related quality of life (QOL) and health economic outcomes, and suggest that achieving inactive disease should be considered a major treatment goal in patients with axial SpA .