Comparative analysis of instruments measuring peripheral arthritis activity in Spondyloarthritis: a measurement properties study in 13 RCTs

Abstract

Objectives

Using randomised controlled trial (RCT) data, this study evaluated the measurement properties of selected instruments to assess peripheral arthritis in axial and peripheral spondyloarthritis (axSpA/pSpA) and to inform selection of the best-performing instrument.

Methods

Eligible studies were RCTs in axSpA or pSpA with individual patient data available to calculate Patient Global Assessment (PGA), swollen/tender joint counts (44 or 66/68 joints), C reactive protein, Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), Axial Spondyloarthritis Disease Activity Score, Disease Activity Index for Psoriatic Arthritis/Disease Activity in Psoriatic Arthritis with 44 joint count (DAPSA44) and 44-joint Disease Activity Score at baseline and timing of primary endpoint. In axSpA, analyses were restricted to patients with peripheral arthritis at baseline. Measurement properties assessed included construct validity, test–retest reliability, longitudinal construct validity, trial discrimination and thresholds of meaning. Risk of bias (RoB) was evaluated and evidence was synthesised using Outcome Measures in Rheumatology-based GREEN/AMBER/RED ratings (low/uncertain/high RoB and good/adequate/poor performance, respectively).

Results

13 RCTs were included (10 axSpA and 3 pSpA). In axSpA trials, 25–51% of patients had peripheral arthritis at baseline (median swollen joint count of 44 joints (SJC44) across studies: 2–3); in pSpA trials, median SJC44 ranged 2–5. In axSpA, construct validity and thresholds of meaning were rated AMBER/GREEN for all instruments, while test–retest reliability was RED only for PGA. Responsiveness was consistently GREEN. Clinical trial discrimination was RED for single-item instruments but consistently GREEN for composite scores. In pSpA, most instruments showed AMBER/GREEN performance, although SJC44 demonstrated RED discrimination.

Conclusions

This study provides the first comprehensive, trial-based evaluation of measurement properties for instruments assessing peripheral arthritis activity in axSpA and pSpA. Composite scores, such as BASDAI, ASDAS and DAPSA44, generally demonstrated more robust performance than single-item instruments, expanding the evidence base for the final instrument-selection phase.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Assessment of peripheral arthritis activity in axial spondyloarthritis (axSpA) and peripheral spondyloarthritis (pSpA) is challenging, and evidence supporting the optimal choice of outcome instruments in clinical trials is limited. To date, the swollen joint count of 44 joints has been the recommended instrument for assessing peripheral arthritis activity, despite demonstrating limited clinical trial discrimination, particularly in axSpA.

WHAT THIS STUDY ADDS

• Using individual patient data from randomised controlled trials, this study provides a comprehensive evaluation of measurement properties of commonly used instruments for peripheral arthritis assessment in axSpA and pSpA. Composite scores consistently outperformed single-item instruments, with Disease Activity in Psoriatic Arthritis with 44 joint count (DAPSA44), Axial Spondyloarthritis Disease Activity Score, Bath Ankylosing Spondylitis Disease Activity Index and 44-joint Disease Activity Score demonstrating the most robust overall performance, including superior clinical trial discrimination. Taking the evidence across measurement properties, DAPSA44 showed the most consistent and adequate performance.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings support the preferential use of composite scores (particularly DAPSA44) for assessing peripheral arthritis activity in SpA clinical trials and inform the Assessment of SpondyloArthritis International Society–Spondyloarthritis and Peripheral Arthritis Disease Activity Instrument Selection and Evaluation instrument-selection process, contributing to more harmonised and evidence-based outcome assessment in future research.

Introduction

Spondyloarthritis (SpA) comprises a heterogeneous group of chronic inflammatory rheumatic diseases characterised by shared clinical, genetic and pathophysiological features, including axial and peripheral musculoskeletal involvement.¹ Based on the predominant pattern of disease expression, SpA is classified into axial spondyloarthritis (axSpA), primarily affecting the spine and sacroiliac joints, and peripheral spondyloarthritis (pSpA), in which peripheral arthritis, enthesitis and dactylitis predominate.^{2 3}

Peripheral arthritis represents a frequent and clinically relevant manifestation in both axSpA and pSpA, contributing substantially to disease burden,^{4 5} functional impairment⁶ and treatment decisions.⁷ However, its assessment remains challenging. In axSpA, despite limited clinical trial discrimination, the swollen joint count of 44 joints (SJC44) is currently recommended in the Assessment of SpondyloArthritis International Society (ASAS) Core Outcome Set (COS) for axSpA⁸ for assessing peripheral arthritis. Importantly, the instrument evaluation preceding the ASAS COS update was restricted to SJC44 and SJC66 and did not consider alternative instruments.

Emerging evidence suggests that composite scores, some originally developed for other inflammatory arthritides, may better capture peripheral arthritis activity. Instruments such as the Disease Activity Index for Psoriatic Arthritis (DAPSA)⁹ or the Disease Activity Score 28,¹⁰ although not routinely applied or formally validated in axSpA or pSpA, have shown promising results,^11–14 highlighting the need for a systematic evaluation of the measurement properties of instruments assessing peripheral arthritis in the context of SpA.

To address this need, the ASAS–SPARADISE (Spondyloarthritis and Peripheral Arthritis Disease Activity Instrument Selection and Evaluation) project was initiated to identify, evaluate and select the most appropriate instrument assessing peripheral arthritis in axSpA and pSpA. The project began with a systematic literature review (SLR) evaluating how peripheral manifestations are assessed in randomised controlled trials (RCTs) of biological and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs). This review demonstrated inconsistent assessment in axSpA trials compared with a more systematic evaluation in pSpA.¹⁵

Following the Outcome Measures in Rheumatology (OMERACT) framework, instruments identified in this initial SLR, together with additional instruments considered clinically relevant by expert consensus, underwent domain match and feasibility evaluation within the ASAS–SPARADISE working group; only those meeting these criteria progressed to the next stage. A subsequent SLR on measurement properties revealed a scarcity of high-quality evidence supporting construct validity, reliability, responsiveness and discrimination of the selected instruments, with composite scores often under-represented or not formally evaluated in SpA populations.¹⁶

To address these gaps, the next stage of ASAS–SPARADISE involved new analyses using individual patient data from RCTs in axSpA and pSpA. Access to multiple trials enabled a comprehensive and methodologically rigorous evaluation of key measurement properties across a range of instruments, including both single-item instruments and composite scores. The present study reports these analyses, aiming to provide evidence to inform instrument selection for peripheral arthritis assessment in clinical research.

Methods

Study design and methodological framework

This study reports new analyses conducted within the framework of the ASAS–SPARADISE project. Overall methodological oversight was provided by the steering committee, composed of two principal investigators, four fellows and three additional expert members. The study followed the principles outlined in the OMERACT Handbook for Instrument Selection for Core Outcome Measurement Sets, which are structured around the pillars of truth, discrimination and feasibility.¹⁷

Study selection and data sources

Eligible studies were RCTs in axSpA or pSpA comparing the effect of b/tsDMARDs versus placebo on peripheral arthritis, for which individual patient data were available to calculate the instruments of interest. Data were accessed through the Vivli data-sharing platform or obtained directly from the principal investigators in case of investigator-initiated studies. Vivli is a global, non-profit platform that provides secure access to anonymised individual participant data from completed RCTs on approved research proposals.¹⁸ Measurement properties were assessed at the individual study level for each instrument, without combining data across trials.

For axSpA trials, analyses were restricted to patients presenting with peripheral arthritis at baseline in order to ensure domain relevance.

Instruments

The following instruments were evaluated across all measurement properties:

• Patient Global Assessment (PGA): Patient-reported outcome assessed with the question ‘How active was your rheumatic disease on average during the last week?’, scored on a 0–10 Numeric Rating Scale (NRS), with higher scores indicating higher disease activity.

• Swollen and tender joint counts: The SJC44/SJC66 and the 44-joint tender joint counts and 68-joint tender joint counts (TJC44/68) assess swelling or tenderness in predefined peripheral joints of the upper and lower extremities (including the temporomandibular, sternoclavicular and acromioclavicular joints when applicable). Each affected joint scores 1 point (range 0 to maximum assessed joints).

• C reactive protein (CRP): Serum CRP (mg/L) was evaluated as an objective inflammatory marker and considered as a standalone instrument.

• Bath Ankylosing Spondylitis Disease Activity Index (BASDAI): Patient-reported outcome addressing fatigue, spinal pain, peripheral joint pain/swelling, enthesitis and morning stiffness (severity and duration), with a final score on a 0–10 NRS, with higher scores indicating higher disease activity.¹⁹

• Axial Spondyloarthritis Disease Activity Score (ASDAS): Composite score combining patient-reported back pain, peripheral pain/swelling, morning stiffness duration, PGA (all 0–10 NRS) and CRP in a weighted formula. Higher scores reflect higher disease activity.²⁰

• Disease Activity Index for Psoriatic Arthritis (DAPSA): Composite score including PGA, patient pain assessment (0–10), SJC66, TJC68 and CRP (mg/dL) calculated as the linear sum of all components with higher values indicating higher disease activity.⁹ When 66-joint counts/68-joint counts were unavailable, a modified version (DAPSA44) was calculated using 44-joint counts converted to 66/68 equivalents using a conversion factor of 1.3: DAPSA44=(1.3×SJC44)+(1.3×TJC44)+PGA+patient pain assessment+high-sensitivity CRP (mg/dL).²¹ If patient pain assessment was unavailable, BASDAI question 3 (joint pain and swelling) was used as a substitute.

• 44-joint Disease Activity Score (DAS44): Composite score originally developed for rheumatoid arthritis, integrating SJC44, TJC44, CRP and PGA using a weighted formula.²²

Detailed computation of each instrument is provided in online supplemental text S1.

Assessment of measurement properties

Following the OMERACT filter V.2.2 for instrument selection,²³ measurement properties from the truth and discrimination pillars were assessed:

Construct validity (truth)

• Construct validity (truth), defined as the extent to which a measurement instrument truly reflects the theoretical concept it intends to measure, was evaluated using two complementary approaches:

1. Percentage fulfilled of hypotheses of the strength of correlation with external constructs: Spearman correlations were calculated between each instrument and measures of related disease domains (Bath Ankylosing Spondylitis Functional Index, ASAS Health Index, Ankylosing Spondylitis Quality of Life, Short Form-36 and EuroQoL). A priori hypotheses regarding expected correlation strength were derived from the ASAS COS exercise and supplemented by the ASAS–perSpA ancillary analysis.^{8 12} Construct validity was determined by the proportion of hypotheses confirmed.

2. Known-group discrimination: Discrimination between groups with different disease activity levels was assessed using standardised mean differences (SMDs). In the absence of a gold standard for peripheral arthritis activity, patients were stratified using PGA (≥5 vs <5) and SJC. Given the high proportion of patients without swollen joints, SJC stratification was performed as ≥1 vs 0 and ≥ vs <median SJC. SMDs were calculated as the difference between group means divided by the pooled SD.

Discrimination

Discrimination encompassed the following measurement properties:

• Test–retest reliability assesses to what extent instrument scores remain unchanged when the construct is (assumed to be) stable. It was assessed using intraclass correlation coefficients (ICC) based on a two-way random-effects model with absolute agreement. As screening data were unavailable for all instruments, reliability was evaluated between baseline and the earliest available of 2-week or 4-week visit in placebo-treated patients, where no meaningful change was expected.

• Longitudinal construct validity (responsiveness) assesses the extent to which an instrument can detect changes in the domain of interest over time (can measure change accurately). It was evaluated using Guyatt’s Responsiveness Index (GRI), standardised response mean (SRM) and effect size (ES), based on change from baseline to the primary endpoint of each trial.

• Clinical trial discrimination assesses the degree to which instruments are sensitive to the related change between the arms of a trial. It was assessed using SMDs of change from baseline, comparing active treatment versus control groups. As no external construct exists, observed discrimination reflects the presence of a treatment effect without specifying the contributing disease domain driving the change, a limitation particularly relevant in axSpA trials, where active axial disease is an inclusion criterion and treatment response is therefore likely to be predominantly driven by the axial component.

To explore whether discrimination improves in the presence of peripheral involvement, SMDs were first calculated in the overall axSpA populations and then recalculated after restricting the analyses to patients with peripheral arthritis at baseline. This second step was performed only for instruments that demonstrated good discrimination in the primary analysis. Results were summarised as the number of trials with improved SMD and the relative change in SMD, both among improved trials and across all trials. Although this approach does not isolate peripheral arthritis activity from overall disease activity, it allows an indirect assessment of whether instruments perform differently in the subgroup in whom peripheral arthritis contributes to the disease burden, and therefore adds to the true discrimination we aim to measure.

• Thresholds of meaning, reflecting disease state using the Patient Acceptable Symptom State (PASS) at the time of the primary endpoint, and meaningful change using those patients who passed from PASS negative at baseline to PASS positive at the time of the endpoint, were explored. PASS is assessed with the question: ‘Considering all the different ways your disease is affecting you, if you were to stay in this state for the next few months, do you consider your current state satisfactory?’. Thresholds were calculated using the 75th percentile method (an alternative to the receiver operating characteristic-based approaches) by identifying the 75th percentile of the score distribution for each instrument.^{24 25} Threshold’s sensitivity and specificity were calculated. For meaningful change, mean differences between groups (patients that changed to PASS positive versus those that remained PASS negative) were compared through Student’s t-test analysis.

Detailed descriptions of all metrics are provided in online supplemental text S2.

Sensitivity analysis

Because DAPSA/DAPSA44 calculations varied across trials depending on the availability of patient pain assessment item, a sensitivity analysis compared the original version (including overall pain) with versions using BASDAI question 3.

Risk-of-bias assessment

Risk-of-bias (RoB) of each study was evaluated separately for each measurement property using the OMERACT Good Methods Checklist.¹⁷ This checklist assesses the appropriateness of the assessment of the measurement property and methodological rigour and categorises results as ‘yes’ (likely low RoB), ‘some cautions’ (can be used as evidence) or ‘no’ (should not be used as evidence).

Evaluation of performance and synthesis of evidence

Quantity, consistency and adequacy of evidence for each measurement property were assessed according to OMERACT recommendations. Construct validity was rated as good when ≥75% of predefined hypotheses were confirmed, adequate when 50–75% and poor when <50%.^{8 23} Correlations were categorised as weak (<0.30), moderate (0.30–0.69) or strong (≥0.70).^{25 26}

Discrimination performance thresholds were defined as follows: SMD≥0.80 indicated good discrimination, 0.50–0.79 adequate and <0.50 poor.^{8 27} Test–retest reliability was considered good when ICC≥0.75, adequate when 0.50–0.74 and poor when <0.50.¹⁷ Longitudinal construct validity (GRI, SRM and ES) was rated using the same SMD cut-offs.^{8 17}

Results were summarised in tables stratified by disease phenotype (axSpA and pSpA) and measurement property. Performance was coded as positive (+) when above the threshold, equivocal (±) when inconsistent or negative (–) when below the threshold. Consistency was defined by consensus based on the proportion of studies demonstrating positive performance for a given property: GREEN when ≥50% of studies were positive, AMBER when <50% and >30% were positive and RED when≤30% were positive.^{14 22}

For thresholds of meaning, given the absence of predefined criteria, performance was considered positive (+) when thresholds were calculated and meaningful change was statistically significant (p<0.05) and equivocal (±) when thresholds were calculated but change was not significant.

Overall synthesis followed OMERACT rules, integrating methodological quality with evidence quantity, consistency and adequacy. Instruments were categorised as GREEN (consistent, adequate or better performance with good methods in ≥2 studies), RED (consistent or questionable inadequate performance), WHITE (no evidence) or AMBER (all other situations).^{16 27} Results were summarised in tables based on the structure and principles of the OMERACT summary of measurement properties framework.²⁸ Guidance for the interpretation of these tables is provided in online supplemental text S3.

All the analyses were performed using Stata SE V.17 and Microsoft Excel.

Results

A total of 13 RCTs were included in the analyses, comprising ten trials in axSpA and three in pSpA. Among the axSpA trials, four included patients with radiographic axSpA (r-axSpA), five focused on non-radiographic axSpA (nr-axSpA) and one included a mixed population (r--axSpA and nr-axSpA). Across axSpA studies, the proportion of patients presenting with peripheral arthritis at baseline ranged from 25% (COAST-V, r-axSpA) to 51% (COAST-X, nr-axSpA). Baseline peripheral joint involvement was generally low, with median SJC44 ranging from 2 to 3 across studies in patients with peripheral arthritis (ie, SJC>0). In pSpA trials, 93–98% of patients presented with peripheral arthritis at baseline, and baseline SJC44 values ranged from 2 to 5 in patients with peripheral arthritis, reflecting a higher degree of peripheral joint involvement compared with axSpA (table 1).

Table 1. Characteristic of the 13 included randomised controlled trials.

Study name	Year	Population	Intervention (vs placebo)	Symptom duration (years)	Patients (n)	Endpoint (weeks)	SJC/TJC assessed	Peripheral arthritis n (%)	Mean (SD) SJC44 in patients with arthritis	Median (IQR) SJC44 in patients with arthritis	Age (years)	Male sex n (%)	HLA-B27 n (%)
Axial spondyloarthritis with peripheral arthritis
ATLAS32	2006	r-axSpA	Adalimumab	10.9 (9.5)*	309	12	44/44	117 (38)	3.9 (4.2)	2 (1–5)	43 (12)	231 (75)	244 (79)
COAST-V33	2018	r-axSpA	Ixekizumab	16.0 (10.2)	167	16	44/44	41 (25)	4.3 (4.3)	2 (2–6)	42 (12)	270 (81)	142 (85)
COAST-W34	2019	r-axSpA	Ixekizumab	19.2 (11.4)	191	16	44/46	82 (43)	4.9 (5.5)	3 (1–6)	47 (13)	159 (83)	158 (82)
SELECT-AXIS35	2019	r-axSpA	Ixekizumab	14.3 (10.7)	186	14	66/68	45 (27)	3.7 (3.4)	3 (2–5)	45 (13)	120 (71)	131 (78)
RAPID-axSpA36	2013	axSpA (r-axSpA+nr-axSpA)	Certolizumab pegol	10.4 (9.5)	318	12	44/44	123 (37)	4.2 (5.3)	2 (1–5)	40 (12)	211 (62)	255 (79)
Ability-137	2013	nr-axSpA	Adalimumab	9.6 (8.7)	184	12	66/68	62 (34)	4.2 (4.3)	2 (1–6)	38 (11)	84 (46)	144 (78)
EMBARK38	2014	nr-axSpA	Etanercept	2.4 (1.9)	212	12	44/44	68 (32)	3.0 (2.9)	2 (1–4)	32 (8)	133 (60)	154 (68)
GO-AHEAD39	2015	nr-axSpA	Golimumab	1.1 (1.3)	188	16	44/44	69 (37)	4.0 (3.2)	3 (2–5)	31 (7)	113 (57)	164 (83)
COAST-X40	2019	nr-axSpA	Ixekizumab	10.8 (9.6)	190	16	44/46	96 (51)	5.7 (6.2)	3 (2–7)	41 (13)	89 (46)	142 (74)
C-AXSPAND41	2019	nr-axSpA	Certolizumab pegol	7.8 (7.4)	281	12	44/44	122 (43)	4.4 (5.8)	3 (2–4)	38 (10)	142 (50)	238 (84)
Peripheral spondyloarthritis
TIPES42	2013	pSpA	Adalimumab	7.3 (7.8)*	40	12	66/68	39 (98)	3.4 (3.3)	2 (1–4)	42 (12)	21 (52)	16 (40)
Ability-243	2015	pSpA	Adalimumab	7.3 (7.2)	161	12	76/78	150 (93)	6.9 (6.6)	5 (2–9)	41 (12)	73 (45)	102 (64)
CRESPA44	2017	pSpA	Golimumab	0.5 (0.2)	60	12	66/68	56 (93)	4.9 (4.7)	4 (2–6)	40 (5)	49 (82)	33 (55)

^*Disease duration.

HLA-B27, Human leukocyte antigen-B27; nr-axSpA, non-radiographic axial Spondyloarthritis; pSpA, peripheral spondyloarthritis; r-axSpA, radiographic axial spondyloarthritis; SJC, swollen joint count; TJC, tender joint count.

Overall methodological quality was high. For all instruments and measurement properties, RoB was rated as GREEN according to the OMERACT Good Methods Checklist. This reflects that all analyses were conducted using appropriate datasets and methodological approaches tailored to each measurement property, in line with OMERACT recommendations.

To illustrate the assessment of measurement properties at the individual study level, the RAPID-axSpA trial is presented as an example in online supplemental tables S1-S4, as it was one of the axSpA studies in which all measurement properties of interest could be evaluated.

Online supplemental table S1, which presents the correlations between instruments and external constructs in RAPID-axSpA, shows the highest construct validity for ASDAS, PGA, SJC and CRP, with up to 100% of a priori hypotheses met. For SJC and CRP, this reflected the fulfilment of hypotheses specifying low expected correlations, while for ASDAS and PGA it reflected the fulfilment of hypotheses specifying mainly moderate correlations with external constructs. Among the remaining instruments, DAPSA44 and BASDAI met 60% of the predefined hypotheses, whereas TJC44 and DAS44 met only 20% and 40%, respectively.

In online supplemental table S2, which assesses construct validity through known-groups discrimination, all instruments except CRP exceeded the threshold (≥0.8), demonstrating good discrimination, whereas CRP showed only adequate discrimination between active and inactive disease.

Test–retest reliability was poor only for PGA (ICC=0.47) and at least adequate for all other instruments, with higher reliability observed for TJC44, CRP and DAS44 (all exceeding the threshold of 0.75) (online supplemental table S3).

Online supplemental table S4 presents results for longitudinal construct validity (responsiveness) and clinical trial discrimination in RAPID-axSpA. All composite scores (BASDAI, ASDAS, DAPSA44 and DAS44), as well as PGA, demonstrated good responsiveness, whereas joint counts (SJC/TJC44) showed poor responsiveness and CRP demonstrated moderate responsiveness. Regarding clinical trial discrimination, ASDAS was the only instrument showing good discrimination between treatment arms; the remaining composite scores and PGA achieved adequate discrimination, while joint counts and CRP showed poor ability to discriminate treatment effects.

Lastly, thresholds of meaning could be established and were judged to be acceptable for all instruments except the joint counts, for which the thresholds for meaningful change between baseline and the study endpoint were not statistically significant (online supplemental tables S10, S11).

Detailed results stratified by measurement property, instrument and individual study are provided in the online supplemental tables S5–S11. Detailed results for the measurement properties for DAPSA44, as well as the synthesis per property, are presented in table 2 as a representative example of the full analytical approach. Results for the remaining instruments are provided in the online supplemental tables S12–S21.

Table 2. Measurement properties for DAPSA44 from 13 randomised controlled trials (10 in axSpA and 3 in pSpA).

Axial spondyloarthritis with peripheral arthritis							Peripheral spondyloarthritis
	Truth-construct validity		Discrimination					Truth-construct validity		Discrimination
RCT name	Strength of correlation	Known-group discrimination	Test–retest reliability	Longitudinal construct validity	Clinical trial discrimination	Thresholds of meaning	RCT name	Strength of correlation	Known-group discrimination	Test–retest reliability	Longitudinal construct validity	Clinical trial discrimination	Thresholds of meaning
ATLAS	+	+	+	+	–		TIPES	+	+		+	+
COAST-V	+	+		+	+		ABILITY-2	–	±	+	+	+	+
COAST-W	+	+		+	–		CRESPA	+	+	+	+	+
SELECT-AXIS	–	+		+	–
RAPID-axSpA	+	+	+	+	+	+
ABILITY-1	–	+		+	+	+
EMBARK	+	+	+	+	–	+
GO-AHEAD	+	+		+	+
COAST-X	+	+		+	–
C-AXSPAND	+	+	+	+	+
Synthesis rating per property	GREEN		GREEN	GREEN	GREEN	GREEN	Synthesis rating per property	GREEN		GREEN	GREEN	GREEN	AMBER

Results in the axSpA RCTs were extracted only from patients with peripheral arthritis at baseline.

‘+’, adequate or good performance; ‘–’, inadequate performance; ‘±’, equivocal performance.

Good methods colour: (light) green, yes, used as evidence; (light) amber, some cautions but used as evidence; (light) red, no, do not use as evidence.

Synthesis rating: GREEN: Good methods used, in at least two pieces of evidence, with consistent findings of adequate or better performance; RED: Good methods used, in at least one or two pieces of evidence, with consistent or questionable findings of inadequate performance; WHITE: no evidence; AMBER: all other situations (not GREEN, RED or WHITE).

ASDAS, Axial Spondyloarthritis Disease Activity Score; axSpA, axial spondyloarthritis; BASDAI, Bath Ankylosing Spondylitis Disease Activity Score; CRP, C reactive protein; DAPSA44, Disease Activity in Psoriatic Arthritis with 44 joint count; DAS44, Disease Activity Score with 44 joint count; PGA, Patient Global Assessment; pSpA, peripheral spondyloarthritis; SJC44, swollen joint count of 44 joints; TCJ44, tender joint count of 44 joints.

Measurement properties in axSpA

In axSpA, construct validity and thresholds of meaning were rated as either AMBER or GREEN for all evaluated instruments, indicating overall adequate to good performance. Test–retest reliability was rated RED only for the PGA, while reliability could not be assessed for the 66-joint/68-joint counts due to a lack of appropriate data (WHITE).

Longitudinal construct validity (responsiveness) showed consistently good performance, with GREEN ratings across instruments. In contrast, clinical trial discrimination demonstrated larger variability. Single-item instruments generally showed poor discrimination between treatment arms and were predominantly rated RED, with the notable exception of CRP, which demonstrated good discrimination and received a GREEN rating. Composite instruments, by contrast, consistently showed high clinical trial discrimination, with GREEN ratings across analyses (table 3).

Table 3. Measurement properties synthesis rating summary per instrument.

Axial spondyloarthritis with peripheral arthritis						Peripheral spondyloarthritis
	Truth	Discrimination					Truth	Discrimination
Instrument	Construct validity	Test–retest reliability	Longitudinal construct validity	Clinical trial discrimination	Thresholds of meaning	Instrument	Construct validity	Test–retest reliability	Longitudinal construct validity	Clinical trial discrimination	Thresholds of meaning
PGA	GREEN	RED	GREEN	RED	GREEN	PGA	GREEN	RED	GREEN	GREEN	AMBER
SJC44	GREEN	GREEN	GREEN	RED	AMBER	SJC44	GREEN	GREEN	GREEN	RED	AMBER
TJC44	AMBER	GREEN	GREEN	RED	AMBER	TJC44	AMBER	AMBER	GREEN	GREEN	AMBER
SJC66	AMBER	WHITE	GREEN	RED	AMBER	SJC66	GREEN	AMBER	GREEN	GREEN	AMBER
TJC68	AMBER	WHITE	GREEN	RED	AMBER	TJC68	GREEN	AMBER	GREEN	GREEN	AMBER
CRP	GREEN	GREEN	GREEN	GREEN	GREEN	CRP	AMBER	AMBER	RED	GREEN	AMBER
BASDAI	GREEN	GREEN	GREEN	GREEN	GREEN	BASDAI	GREEN	AMBER	GREEN	GREEN	AMBER
ASDAS	GREEN	GREEN	GREEN	GREEN	GREEN	ASDAS	GREEN	AMBER	GREEN	GREEN	AMBER
DAPSA	AMBER	WHITE	GREEN	AMBER	AMBER	DAPSA	GREEN	AMBER	GREEN	GREEN	AMBER
DAPSA44	GREEN	GREEN	GREEN	GREEN	GREEN	DAPSA44	GREEN	AMBER	GREEN	GREEN	AMBER
DAS44	AMBER	GREEN	GREEN	GREEN	AMBER	DAS44	GREEN	AMBER	GREEN	GREEN	AMBER

Data from 13 randomised controlled trials (10 in axSpA and 3 in pSpA).

Results in the axSpA RCTs were extracted only from patients with peripheral arthritis at baseline.

‘+’, adequate or good performance; ‘−’, inadequate performance; ‘±’, equivocal performance.

Good methods colour: (light) green, yes, used as evidence; (light) amber, some cautions but used as evidence; (light) red, no, do not use as evidence.

Synthesis rating: GREEN: Good methods used, in at least two pieces of evidence, with consistent findings of adequate or better performance; RED: Good methods used, in at least one or two pieces of evidence, with consistent or questionable findings of inadequate performance; WHITE: no evidence; AMBER: all other situations (not GREEN, RED or WHITE).

ASDAS, Axial Spondyloarthritis Disease Activity Score; axSpA, axial spondyloarthritis; BASDAI, Bath Ankylosing Spondylitis Disease Activity Score; CRP, C reactive protein; DAPSA44, Disease Activity in Psoriatic Arthritis with 44 joint count; DAS44, Disease Activity Score with 44 joint count; PGA, Patient Global Assessment; pSpA, peripheral spondyloarthritis; SJC44, swollen joint count of 44 joints; TCJ44, tender joint count of 44 joints.

Further analyses exploring the influence of axial component on clinical trial discrimination are presented in online supplemental table S22. Among instruments with adequate or good discrimination (CRP, BASDAI, ASDAS, DAPSA44 and DAS44), improvements in SMD when restricting analyses to axSpA patients with peripheral arthritis were more frequent for instruments including joint counts (DAPSA44 and DAS44) and for CRP. Improvements in SMD were observed in four trials for BASDAI and ASDAS, six trials for both DAPSA44 and DAS44 and seven for CRP. When restricting the analysis to trials showing improvement, the relative increase in SMD was 22% for CRP, 50% for BASDAI, 23% for ASDAS, 53% for DAPSA44 and 81% for DAS44. When all trials were considered, including those without improvement, relative changes were 11% for CRP, 2% for BASDAI, −1% for ASDAS, 16% for DAPSA44 and 51% for DAS44 (online supplemental table S22).

Lastly, thresholds of meaning were rated AMBER for joint counts and DAPSA and GREEN for the remaining instruments (table 3).

Measurement properties in pSpA

In pSpA trials, construct validity and test–retest reliability were rated AMBER to GREEN for all instruments, with the exception of PGA, for which reliability was poor (RED). Longitudinal construct validity was consistently rated GREEN, except for CRP, which showed poor performance (RED). A similar pattern was observed for clinical trial discrimination, with all instruments rated GREEN except for the SJC44, which demonstrated poor discrimination (RED). Thresholds of meaning were rated AMBER for all instruments (table 3).

Sensitivity analysis

Correlations with external constructs were highly similar between homologous versions of DAPSA and DAPSA44 using BASDAI question 3 versus patient pain assessment, with absolute differences≤0.05 across all correlations (online supplemental table S22). SMDs for known-groups discrimination were virtually identical between score versions (online supplemental table S23). Comparable results were observed for test–retest reliability (online supplemental table S24), longitudinal construct validity and clinical trial discrimination (online supplemental table S25), as well as for thresholds of meaning (online supplemental table S26), with no meaningful differences between the alternative and original versions.

Discussion

Our study provides the first comprehensive evaluation of measurement properties of instruments used to assess peripheral arthritis activity in RCTs of axSpA and pSpA. Building on evidence gaps identified in previous ASAS–SPARADISE SLRs,¹⁵ these new analyses using individual patient data from multiple trials generate robust and methodologically consistent evidence of measurement properties for all the instruments of interest.

A central finding is the clear contrast between single-item instruments and multiple-item instruments (composite scores), particularly in axSpA. While joint counts and PGA demonstrated acceptable construct validity and responsiveness, they consistently showed limited clinical trial discrimination. This was most evident for SJC44, currently recommended in the ASAS COS for axSpA.⁸ Notably, during the ASAS COS development, clinical trial discrimination of SJC44 was already identified as poor (based on data from two RCTs); however, no better-supported alternatives were available, and it was retained. By substantially expanding the available evidence, the present study confirms that SJC44 lacks adequate sensitivity to discriminate treatment effects.

In contrast, composite scores (DAPSA, DAPSA44, ASDAS, DAS44 and BASDAI) demonstrated robust construct validity, strong longitudinal construct validity and good clinical trial discrimination in axSpA. By integrating PROs, joint counts and objective inflammatory markers, these instruments appear to capture peripheral arthritis activity more comprehensively than single-item measures, resulting in more reliable and responsive to change instruments.^{29 30}

In pSpA, most instruments (except SJC44) demonstrated acceptable performance across measurement properties, including clinical trial discrimination, likely reflecting the higher prevalence of peripheral arthritis. However, discrimination was generally adequate (AMBER) rather than consistently strong. Composite indices, by contrast, demonstrated higher discrimination, more frequently at good level (GREEN), with larger SMDs overall. Notably, the good performance of ASDAS and BASDAI was consistent with previous findings in pSpA, where both indices demonstrated good construct validity and responsiveness despite not incorporating joint counts.¹³ Moreover, the poor discrimination observed for SJC44 in pSpA highlights that joint counts alone may be insufficient as a primary trial endpoint, even in predominantly peripheral disease.

From a methodological perspective, interpretation of clinical trial discrimination in axSpA remains complex, as treatment effects may reflect improvements across different disease domains. Although primary analyses were restricted to patients with peripheral arthritis, axial improvement may still contribute due to trial inclusion criteria.

To explore this, we compared discrimination in overall axSpA populations versus subgroups with peripheral arthritis. Instruments incorporating joint counts (DAPSA44 and DAS44), more frequently demonstrated improved discrimination and larger relative SMD increases than indices primarily driven by axial disease (BASDAI and ASDAS). Although CRP improved in several trials, relative changes were more modest. Because all composite instruments reflect overall disease activity rather than peripheral arthritis in isolation, these findings should be interpreted as indirect evidence that instruments incorporating joint counts are more sensitive to treatment effects in patients with peripheral arthritis.

Another important methodological consideration was the application of uniform performance criteria across instruments to support the OMERACT-based synthesis ratings. To this end, identical a priori hypotheses with a minimum performance threshold were defined for all instruments for each measurement property, with the requirement that instruments demonstrate at least adequate performance. This ensured that instruments met meaningful performance standards rather than minimal statistical criteria.

Comparing multiple instruments within the same domain posed an additional challenge, as OMERACT primarily supports the evaluation of individual measures. To address this, we applied a structured approach allowing transparent comparison across measurement properties across instruments while remaining aligned with OMERACT principles.

Several limitations should be acknowledged. First, axSpA analyses were restricted to patients with peripheral arthritis at baseline. Although this restriction was necessary to ensure domain relevance, it reduced sample sizes and may have limited the robustness of the results. Second, since screening data were unavailable for all instruments, test–retest reliability was estimated using baseline and early placebo visits, assuming stability. Nevertheless, a placebo effect cannot be fully excluded. Third, thresholds of meaning were calculable in few studies, resulting in limited evidence and predominantly AMBER ratings. Finally, evidence in pSpA was particularly limited due to the small number of existing RCTs, which may restrict the certainty and generalisability of findings in this population.

Despite these limitations, the strengths of this study are considerable. The use of individual patient data from multiple RCTs, the application of predefined hypotheses and standardised performance criteria, and the integration of methodological quality assessment provide a level of evidence that has been largely absent from the field to date. An additional strength of the present analyses is supported by the sensitivity analyses comparing the original DAPSA and DAPSA44 scores with alternative versions in which BASDAI question 3 was used in place of the patient pain assessment item. Although BASDAI question 3 was consistently used across all trials to ensure methodological consistency, sensitivity analyses confirmed that this approach did not alter the measurement performance of the original scores. Across all evaluated measurement properties, performance was highly comparable between original and alternative score versions. These findings support the use of BASDAI question 3 as a suitable substitute when the patient pain assessment item is not available, without compromising the measurement performance of DAPSA or DAPSA44. It is important to note that the present study was not designed to evaluate the interchangeability of DAPSA44 and the original DAPSA, nor to determine whether one instrument should replace the other. Rather, both instruments were evaluated independently with respect to their measurement properties using RCTs. Therefore, the present findings should not be interpreted as evidence supporting interchangeability between these instruments. Dedicated validation studies specifically designed to address this question in PsA, axSpA and pSpA have been performed separately.³¹

To summarise, the present analyses provide robust evidence supporting the use of composite scores for the assessment of peripheral arthritis activity in axSpA and pSpA clinical trials. In particular, composite scores that explicitly incorporate joint counts, such as DAPSA44 and DAS44, demonstrated larger improvement in clinical trial discrimination when comparing performance between the overall axSpA population versus the axSpA with peripheral arthritis population. While single-item instruments such as joint counts remain valuable for descriptive and clinical purposes, their limited ability to discriminate treatment effects indicates that they should not be used in isolation in clinical trial settings. Taking the overall body of evidence across measurement properties, DAPSA44 showed the most consistent and adequate performance.

Together, these findings directly inform the ASAS–SPARADISE instrument-selection process and support the preferential use of composite scores (BASDAI, ASDAS, DAPSA44 and DAS44) for harmonised, evidence-based assessment of peripheral arthritis in SpA clinical research.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.