Assessing vasculitis in Takayasu arteritis by ultrasound: results of OMERACT patient-based reliability exercises

Abstract

Objective

To evaluate the inter-rater and intrarater reliability of consensus-based definitions of elementary ultrasound lesions in Takayasu arteritis (TAK) in live patient-based reliability exercises.

Methods

Two live exercises with two rounds each were conducted, each involving 12 vascular sonographers who independently examined five patients with TAK and one control subject. Bilateral common carotid, subclavian and axillary arteries and the abdominal aorta were investigated to identify key elementary ultrasound lesions in TAK: ‘macaroni sign’, stenosis and occlusion. Inter-rater and intrarater reliability were calculated using Fleiss’ and Cohen’s kappa (κ), respectively.

Results

Overall, inter-rater reliability across the two rounds of both exercises was moderate to good for the ‘macaroni sign’ (κ 0.51-0.74 across the two rounds of both exercises), fair for stenosis (0.24–0.38) and variable for occlusion (0.02–0.60). Intrarater reliability across the two exercises was good to excellent for the ‘macaroni sign’ (0.74–0.81) and moderate for stenosis (0.55–0.64) and occlusion (0.64–1.00; the latter based on a limited number of assessments). In both exercises, inter-rater reliability was good to excellent for the ‘macaroni sign’ in carotids but worse across other vessels and lesion types; intrarater reliability was good to excellent when analysed per single vessel and lesion though limited for occlusions and for the abdominal aorta.

Conclusion

Among the ultrasound elementary lesions for TAK, the ‘macaroni sign’ revealed moderate to good reliability, while the reproducibility of stenosis and occlusion was worse and more variable. These results provide the methodological basis to develop a new global Outcome Measures in Rheumatology (OMERACT) ultrasound score in TAK.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Ultrasound allows direct, non-invasive assessment of vessel wall and luminal changes in Takayasu arteritis (TAK) but lacks standardisation as an outcome measure. Within Outcome Measures in Rheumatology (OMERACT), consensus-based definitions of key lesions ('macaroni sign’, stenosis, occlusion) and core arterial territories showed promising reliability in online image-based exercises, yet patient-based reproducibility had not been established.

WHAT THIS STUDY ADDS

• In live patient-based exercises, the ‘macaroni sign’ demonstrated moderate to good inter-rater and good to excellent intrarater reliability, particularly in carotid and axillary arteries, whereas stenosis and occlusion showed more variable agreement. Structured calibration between sessions significantly improved reproducibility.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings support vessel wall thickening as the most reliable ultrasound feature in TAK, provide the empirical basis for a composite OMERACT ultrasound score and advance ultrasound qualification as an outcome measure in clinical trials, while underscoring the importance of harmonised protocols and training for consistent multicentre assessments.

Introduction

Takayasu arteritis (TAK) is a chronic large-vessel vasculitis affecting the aorta and its major branches, with clinical manifestations ranging from non-specific constitutional symptoms to ischaemic complications such as stenosis or occlusion.¹ Although C reactive protein and erythrocyte sedimentation rate are established blood biomarkers for monitoring disease activity, they may remain normal in certain patients despite disease progression, and they can be affected directly by targeted treatments.^{1 2} In these cases, disease monitoring requires integration of clinical evaluation with imaging, which plays an increasing role both in assessing disease activity and in evaluating treatment response.

Several imaging modalities are currently applied for monitoring TAK, including magnetic resonance angiography, CT angiography, 18F-fluorodeoxyglucose positron emission tomography and ultrasound.³ Each technique has specific advantages and limitations, and their use is often determined by availability, costs, local expertise, potential impact of recurrent radiation exposure and clinical context. Although ultrasound has some limitations as a diagnostic tool due to limited visibility of certain arterial segments such as the thoracic aorta,⁴ it represents an attractive option for longitudinal monitoring of patients, as it is non-invasive, radiation-free, widely available and cheaper than the other imaging modalities. In TAK, ultrasound allows direct visualisation of the vessel wall thickness and luminal abnormalities, thereby providing valuable information on both inflammatory and structural changes.

Although imaging is increasingly used in clinical practice, it has not yet been incorporated as a standardised outcome measure in clinical trials of TAK; nevertheless, some randomised controlled trials have reported imaging findings as secondary or exploratory endpoints.⁵ The Large-Vessel Vasculitis subgroup of the OMERACT Ultrasound Working Group aims to develop standardised ultrasound scores that could serve as outcome measures for future studies in vasculitis, enabling harmonised and reproducible assessments of disease activity and treatment response according to the OMERACT Instrument Selection Algorithm (OFISA).^6–8 As part of the project to develop a new ultrasound score for TAK to be used for this purpose, we conducted a systematic literature review (SLR) followed by a Delphi process, which led to consensus-based definitions of the following elementary ultrasound lesions in TAK: ‘macaroni sign’, stenosis and occlusion.^{9 10} During this process, it was agreed to focus on a core set of arterial territories considered most informative and accessible for ultrasound evaluation: bilateral common carotid, subclavian and axillary arteries as well as the abdominal aorta. These definitions were subsequently tested in an online reliability exercise using static images, which showed promising results.¹⁰

According to the OMERACT framework, the next step in the process of developing a new outcome measure (but not a diagnostic tool) is to test the candidate definitions in live exercise(s) involving patients.⁸ The present study was, therefore, conducted to evaluate the inter-rater and intra-rater reliability of the new consensus-based definitions when applied to standardised ultrasound assessments of patients with TAK and controls.

Methods

Study design and setting

Two live reliability exercises were conducted in October 2024 and May 2025 in Innsbruck, Austria. Each exercise spanned 2 days and involved experienced vascular sonographers from the OMERACT Large-Vessel Vasculitis Ultrasound subgroup performing standardised ultrasound assessments on five TAK patients and one control subject. The sample size, including one healthy control, was informed by earlier OMERACT ultrasound patient-based reliability exercises¹¹ and by feasibility considerations inherent to live, patient-based OMERACT reliability studies but was not subject to formal sample size calculation. The inclusion of a healthy control was not intended to test the diagnostic accuracy of sonography, but to ensure that none of the consensus-based elementary ultrasound lesions would be detected in non-vasculitic arterial segments. The examined subjects were selected by two convenors (AT, CDu), who did not participate in the reliability assessment. All participants provided written informed consent, and the study was approved by the local ethics committee (1197/2024).

On day 1 of the first exercise, a theoretical training session (2.5 hours) was held to review the study objectives, scanning protocol, ultrasound settings and consensus-based definitions of elementary ultrasound lesions and arterial segments. During the second exercise, this session was expanded to allow more time (4 hours) for hands-on calibration and a more extensive alignment on ultrasound machine settings. In addition, discrepancies observed during the first exercise were discussed in detail, leading to improved clarity in the interpretation of the definition of lesions. This session was further complemented by practical scanning on one patient with TAK (not included in the reliability assessments) to better align experts with the scanning protocol.

On day 2 of each exercise, all sonographers independently examined six individuals (five TAK patients and one healthy control) in a random order. The same subjects were rescanned approximately 3 hours later in a different order to assess intrarater reliability.

The rationale for conducting a second exercise was consistent with a stepwise, iterative approach to outcome measure development within OMERACT and aimed to evaluate whether additional calibration and practical alignment could improve agreement across readers.

Ultrasound examination

Sonographers assessed seven predefined arterial segments: bilateral common carotid, subclavian and axillary arteries and the abdominal aorta. For each segment, the presence or absence of the three elementary ultrasound lesions (‘macaroni sign’, stenosis, occlusion) was recorded. Based on the results of the SLR and Delphi consensus,^{9 10} sonographers did not assess structural damage (ie, aneurysms or dissections) or describe atherosclerotic plaques.

To ensure standardised and comprehensive assessment, sonographers were given precise instructions on the evaluation of each arterial segment (table 1). On completing the ultrasound examination, sonographers provided a judgement on the presence or absence of TAK.

Table 1. Consensus-based definitions of elementary ultrasound lesions in Takayasu arteritis (TAK)¹⁰ and sonographer instructions to ensure a standardised assessment of the arterial segments evaluated in the reliability exercise.

Definitions of elementary ultrasound lesions in TAK
Macaroni sign	Homogeneous, usually hypo-echoic to iso-echoic wall thickening, well delineated towards the luminal side, visible both in longitudinal and transverse planes, most commonly concentric in transverse scans.
Stenosis	Usually hypo–isoechoic vasculitic vessel wall thickening associated with characteristic Doppler curves showing turbulence (aliasing and/or persistent diastolic flow by colour Doppler) and increased systolic and diastolic blood flow velocities.
Occlusion	Absence of colour Doppler signals in a visible artery, filled with usually hypoechoic material even with low pulse repetition frequency and high colour gain.
Arterial segments assessed during the exercise and vessel-specific instructions used to ensure standardised assessment.
All vessels	Evaluated both in longitudinal and transverse scans
Carotid arteries	Assessment is started above the clavicle, where the transducer can be positioned perpendicular to the vessel
Subclavian arteries	Assessed avoiding placing the transducer too deep below the clavicle
Axillary arteries	Assessed in all three segments: a) from the anterior side: to assess the proximal segment (starting below the clavicle) b) from the axillary side: to assess the middle and distal segments
Abdominal aorta	Assessment of the distal abdominal aorta, defined as the sub-renal segment extending to the bifurcation of the common iliac arteries.

All examinations were performed with the study subjects in a supine position. Sonographers were allocated a total of 20 min per subject (18 min for scanning and 2 min for rotation). Each study subject was assigned to a dedicated ultrasound station and remained in place to ensure consistency, while sonographers rotated between stations and consequently used different ultrasound devices throughout the exercise. For both exercises, a first round was completed in the morning and a second in the afternoon to assess intrarater reliability; before the second round, study subjects were randomly redistributed among ultrasound machines. Examiners were blinded to diagnosis (TAK vs control), all clinical data and the presence or absence of lesions in each arterial segment. They were not permitted to communicate with each other nor with the study subjects during the exercise. To facilitate standardised data collection, each sonographer was assisted by a medical student who recorded findings in real time on paper-based case report forms, which were subsequently transferred into an electronic database.

Definitions

The following elementary ultrasound lesions were assessed in all arterial segments, based on consensus definitions developed through prior SLR and Delphi process¹⁰: (1) ‘macaroni sign’, (2) stenosis and (3) occlusion. Definitions of these lesions as agreed in the previous Delphi exercise are detailed in table 1.

Ultrasound equipment and setting

In both exercises, ultrasound examinations were performed using six machines provided by Canon Medical Systems (Austria). The first exercise involved five Canon Aplio A and one Canon Aplio Me system, equipped with PLT-705BT (frequency range 3.0–8.5 MHz) and PLU-805BT (3.0–12.0 MHz) linear transducers, respectively. The second exercise involved one Aplio A system equipped with a PLT-705BT linear transducer (3.0–8.5 MHz), two Aplio Me systems equipped with PLU-805BT linear transducers (3.0–12.0 MHz), one Aplio Beyond system equipped with a PLU-805BT linear transducer (3.0–12.0 MHz), one Aplio i800 system equipped with a PLT-705BX linear transducer (3.5–11.0 MHz) and one Aplio Flex system equipped with a PLC-1004LT linear transducer (4.4–14.0 MHz) to assess the carotid arteries and a PLC-704LT linear transducer (3.0–11.0 MHz) to assess the remaining vascular territories. All ultrasound systems were operated in accordance with predefined settings (see details in online supplemental table 1) optimised by experts with the assistance of experienced ultrasound technicians. When needed, PRF, gain and depth were further adjusted by the sonographers during the scan to achieve the best image quality.

Statistical analysis

We conducted descriptive statistics reporting absolute numbers and percentages for categorical variables, and mean±SD or median and IQR for continuous variables, as appropriate. Inter-rater and intrarater reliability were assessed for each elementary lesion (‘macaroni sign’, stenosis, occlusion) overall and separately for each vascular territory. Inter-rater and intrarater reliability were also assessed for the presence/absence of TAK. Inter-rater reliability was calculated using Fleiss’ kappa (κ) for multiple raters. Intrarater reliability was assessed using Cohen’s κ for each sonographer, and the mean and SD of individual κ were computed. Kappa values were interpreted according to Landis and Koch: values ≤0.20 indicate poor agreement, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 good and ≥0.81 excellent agreement. In addition, the proportion of identical ratings among observers was calculated for both inter-rater and intrarater assessments. All statistical analyses were performed using R Statistical Software (R Foundation for Statistical Computing, Vienna, Austria).

Results

Study population and sonographers

Five patients with TAK participated in each exercise, two patients from the first exercise were also included in the second. All patients fulfilled the 2022 American College of Rheumatology/EULAR classification criteria.¹² In the first exercise, patients had a mean age of 28.6±11.2 years and a mean disease duration of 64.6±51.7 months; in the second exercise, patients were slightly younger (mean age 24.2±4.7 years), with longer disease duration (74.8±73.8 months). Most patients were female (60% in exercise 1; 100% in exercise 2). Only one patient, in the first exercise, had active disease at the time of the ultrasound evaluation according to the managing rheumatologist. Most patients received glucocorticoid therapy, and all were on glucocorticoid-sparing agents (online supplemental table 2). Each exercise also included one healthy subject.

Twelve vascular sonographers participated in each exercise. Eight experts were involved in both sessions, while four were replaced in the second exercise due to unavailability. In the first exercise, the mean age of the sonographers was 47.9±9.5 years, and the mean experience in vascular ultrasound was 14.3±7.0 years; 4 of the 12 sonographers were women. In the second exercise, the mean age was 48.8±8.9 years, and the mean experience was 15.9±9.5 years, with five women among the 12 sonographers (online supplemental table 3).

Patient-based reliability exercises

Across both exercises, the ‘macaroni sign’ was the most frequently detected lesion, followed by stenosis and, more rarely, occlusion (table 2).

Table 2. Detection rates of ultrasound elementary lesions by 12 expert sonographers in 2 reliability exercises on Takayasu arteritis (TAK) patients (n=5) and healthy controls (n=1).

	Carotid arteries (n=12)	Subclavian arteries (n=12)	Axillary arteries (n=12)	Abdominal aorta (n=6)	Overall (n=42)
Exercise 1, number of positive findings/vessel segments (range)
Round 1
Macaroni sign	7 (5–9)	2.5 (0–8)	2.5 (0–4)	0 (0–5)	12.5 (5–25)
Stenosis	3 (1–7)	1 (0–5)	1 (0–3)	0 (0–3)	5 (2–12)
Occlusion	0 (0–2)	1 (0–2)	0 (0–1)	0 (0)	2 (0–3)
Round 2
Macaroni sign	7 (5–9)	2 (0–6)	3 (0–5)	0.5 (0–4)	12.5 (7–19)
Stenosis	3.5 (1–6)	1 (0–9)	1 (0–3)	0 (0–4)	6 (4–19)
Occlusion	0 (0–1)	0 (0–2)	0 (0–1)	0 (0–1)	1 (0–2)
Exercise 2, number of positive findings/vessel segments (range)
Round 1
Macaroni sign	7 (4–7)	3 (2–7)	2.5 (1–5)	0 (0–3)	13 (7–19)
Stenosis	2 (0–3)	2 (0–5)	1 (0–2)	0 (0–1)	4.5 (2–9)
Occlusion	0 (0–2)	0.5 (0–2)	0 (0)	0 (0)	1 (0–2)
Round 2
Macaroni sign	7 (4–8)	2 (0–6)	3 (1–4)	0 (0–2)	12.5 (8–17)
Stenosis	2 (0–5)	1.5 (0–4)	1.5 (0–4)	0 (0)	5.5 (1–9)
Occlusion	0 (0)	1 (0–2)	0 (0)	0 (0)	1 (0–2)

Data are expressed as median (range), n, number.

In the first exercise, overall inter-rater reliability for the ‘macaroni sign’ was moderate (κ=0.51 in round 1 and 0.54 in round 2), improving to good in the second (κ=0.74 in both rounds). Overall agreement for stenosis was fair in both exercises (κ=0.24–0.26 in exercise 1 and 0.34–0.38 in exercise 2). Reliability for occlusion was poor in the first exercise (κ=0.02–0.05) but increased markedly in the second exercise (κ=0.11–0.60), reaching good levels in round 2. Inter-rater agreement for the presence/absence of TAK improved from moderate in the first exercise (κ=0.44–0.48) to excellent in the second (κ=0.94–1.00) (table 3).

Table 3. Inter-rater reliability (Fleiss’ κ) for ultrasound elementary lesions in Takayasu arteritis (TAK) patients (n=5) and healthy controls (n=1), assessed by 12 expert sonographers.

	Fleiss’ κ (95% CI)
	Round 1	Round 2
Exercise 1
Overall
Macaroni sign	0.51 (0.47 to 0.55)	0.54 (0.50 to 0.58)
Stenosis	0.26 (0.22 to 0.30)	0.24 (0.20 to 0.28)
Occlusion	0.05 (0 to 0.1)	0.02 (0 to 0.07)
Presence/absence of TAK	0.44 (0.34 to 0.54)	0.48 (0.39 to 0.58)
Macaroni sign
Carotid artery	0.67 (0.60 to 0.74)	0.68 (0.61 to 0.75)
Subclavian artery	0.02 (0 to 0.09)	0.30 (0.22 to 0.37)
Axillary artery	0.26 (0.19 to 0.34)	0.40 (0.33 to 0.47)
Abdominal aorta	0 (0 to 0.12)	0.17 (0.06 to 0.28)
Stenosis
Carotid artery	0.30 (0.23 to 0.37)	0.23 (0.16 to 0.30)
Subclavian artery	0.04 (0 to 0.12)	0.22 (0.15 to 0.30)
Axillary artery	0.03 (0 to 0.10)	0.25 (0.19 to 0.32)
Abdominal aorta	0 (0 to 0.11)	0 (0 to 0.08)
Occlusion
Carotid artery	0.04 (0 to 0.11)	0.07 (0 to 0.15)
Subclavian artery	0.03 (0 to 0.11)	0 (0 to 0.07)
Axillary artery	n.a.*	0 (0 to 0.06)
Abdominal aorta	n.a.*	n.a.*
Exercise 2
Overall
Macaroni sign	0.74 (0.70 to 0.78)	0.74 (0.70 to 0.78)
Stenosis	0.34 (0.30 to 0.38)	0.38 (0.34 to 0.42)
Occlusion	0.11 (0.10 to 0.15)	0.60 (0.56 to 0.64)
Presence/absence of TAK	0.94 (0.84 to 1.0)	1.0 (0.9 to 1.0)
Macaroni sign
Carotid artery	0.88 (0.81 to 0.95)	0.82 (0.75 to 0.89)
Subclavian artery	0.55 (0.48 to 0.63)	0.45 (0.38 to 0.52)
Axillary artery	0.64 (0.57 to 0.72)	0.74 (0.66 to 0.81)
Abdominal aorta	0 (0 to 0.12)	0 (0 to 0.11)
Stenosis
Carotid artery	0.34 (0.27 to 0.41)	0.36 (0.29 to 0.43)
Subclavian artery	0.28 (0.20 to 0.36)	0.39 (0.32 to 0.47)
Axillary artery	0.38 (0.30 to 0.45)	0.36 (0.29 to 0.43)
Abdominal aorta	0 (0 to 0.10)	n.a.*
Occlusion
Carotid artery	0.04 (0 to 0.11)	n.a.*
Subclavian artery	0.16 (0.09 to 0.24)	0.58 (0.51 to 0.65)
Axillary artery	n.a.*	n.a.*
Abdominal aorta	n.a.*	n.a.*

Carotid, subclavian and axillary arteries have been investigated bilaterally.

^*Fleiss κ cannot be calculated because all values are identical.

When analysed by vascular territory, inter-rater reliability was consistently highest in the carotid arteries and lowest in the abdominal aorta, with intermediate and more variable agreement in the subclavian and axillary arteries. When comparing the first and second exercise, the greatest improvement was observed for the ‘macaroni sign’ (rising from moderate to good/excellent particularly in the carotid, subclavian and axillary arteries) and for occlusion (mainly in the subclavian arteries) (table 3).

The main sources of disagreement identified during the first exercise and the corresponding clarifications agreed on before the second exercise are summarised in online supplemental table 4.

Intrarater reliability was good to excellent for the ‘macaroni sign’ in both exercises, increasing from a mean κ of 0.74±0.10 in the first to 0.81±0.11 in the second. Agreement for stenosis was good in the first (0.64±0.17) and moderate in the second (0.55±0.16). For occlusion, κ values were more variable because of the limited number of evaluable segments but showed a tendency toward higher consistency in the second exercise (0.64±0.22; 1.0±0.0) (table 4).

Table 4. Intrarater agreement (Cohen’s κ) for ultrasound elementary lesions in Takayasu arteritis (TAK) patients (n=5) and healthy controls (n=1) assessed by 12 expert sonographers.

	Exercise 1		Exercise 2
		Mean (SD)		Mean (SD)
Overall
Macaroni sign	(n=12)	0.74 (0.10)	(n=12)	0.81 (0.11)
Stenosis	(n=12)	0.64 (0.17)	(n=12)	0.55 (0.16)
Occlusion	(n=6)*	0.64 (0.22)	(n=5)*	1.0 (0)
Presence/absence of TAK	(n=12)	0.87 (0.19)	(n=12)	0.96 (0.12)
Macaroni sign
Carotid artery	(n=12)	0.91 (0.12)	(n=12)	0.94 (0.11)
Subclavian artery	(n=9)*	0.59 (0.33)	(n=11)*	0.63 (0.19)
Axillary artery	(n=9)*	0.64 (0.32)	(n=12)	0.79 (0.28)
Abdominal aorta	(n=3)*	0.51 (0.17)	(n=3)*	1.0 (0)
Stenosis
Carotid artery	(n=12)	0.65 (0.34)	(n=9)*	0.75 (0.25)
Subclavian artery	(n=7)*	0.68 (0.38)	(n=9)*	0.62 (0.22)
Axillary artery	(n=6)*	0.58 (0.44)	(n=8)*	0.54 (0.36)
Abdominal aorta	(n=2)*	1.0 (0)	(n=0)*	n.a.
Occlusion
Carotid artery	(n=2)*	1 (0)	(n=0)*	n.a.
Subclavian artery	(n=4)*	0.91 (0.19)	(n=5)*	1.0 (0)
Axillary artery	(n=1)*	0	(n=0)*	n.a.
Abdominal aorta	(n=0)*	n.a.	(n=0)*	n.a.

Carotid, subclavian and axillary arteries have been investigated bilaterally.

^*κ values could not be calculated for all investigators, because at least one of the values in the 4x4 table was a constant.

Across vascular territories, intrarater agreement was again highest in the carotid and axillary arteries, and worst in the abdominal aorta. In both exercises, intrarater reliability for the presence/absence of TAK was excellent, with κ values approaching 1.0 (0.87±0.19; 0.96±0.12) (table 4).

Discussion

In these live patient-based reliability exercises of ultrasound elementary lesions in TAK, conducted as part of an OMERACT initiative to develop an ultrasound composite score for monitoring/outcome measure, we observed moderate to good inter-rater and intrarater reliabilities for the ‘macaroni sign’, particularly at the carotid and axillary arteries, while agreement for stenosis and occlusion was more variable. Ultrasound examination also demonstrated good overall reliability for the presence/absence of TAK.

Two live reliability exercises were conducted to refine the assessment of the consensus definitions previously tested in an online study on static images.¹⁰ The second exercise revealed higher κ values than the first, suggesting that the expanded calibration, with additional time devoted to protocol alignment, more consistent harmonisation of machine settings, focused discussion of discrepancies in the first round and practical scanning of a TAK patient, likely contributed to improving agreement between experts. Compared with the previous online study,¹⁰ reliability in the onsite exercises was lower; however, onsite assessment allowed for a more realistic estimate of sonographers’ variability by incorporating all technical aspects of image acquisition, which are crucial for ultrasound.⁸ These observations also mirror previous OMERACT experiences, particularly in GCA, where online reliability exercises resulted in higher κ values than onsite studies, and where repeated calibration and live testing led to a marked improvement of the reliability of elementary lesions.¹¹

While the ‘macaroni sign’ performed best in terms of reliability, there was still some heterogeneity according to the arterial segment investigated. This was most likely influenced by the different prevalence of the ‘macaroni sign’ at individual vessels (and the dependency of κ statistics on the proportion of positive and negative results as explained below) as well as by differences in accessibility of the examined arteries. Visualisation of subclavian and proximal axillary arteries, for example, can be technically challenging.¹³ In addition, the abdominal aorta is often obscured by intestinal gas, limiting the image quality.⁴

Similar to GCA, arterial wall thickening appears to be the most reliable ultrasound correlate of vascular involvement in TAK.^14–18 In contrast, haemodynamic lesions such as stenosis and occlusion exhibited greater variability in terms of κ values. Several factors may account for this observation: first, Doppler assessments depend not only on vessel morphology but also on blood flow conditions, insonation angle and operator interpretation of spectral and colour Doppler.¹⁹ Moreover, no standardised cut-off values for pulsed-wave Doppler velocities have been established for the vessels investigated, except for the carotid arteries, thereby adding subjectivity to the interpretation of the exam. Second, the low prevalence of advanced luminal lesions in our cohort, particularly occlusions, limited the number of positive findings available for evaluation, which in turn affects the robustness of the corresponding κ values. Indeed, κ statistic performs best when both positive and negative cases are equally represented.²⁰ These methodological considerations underline the importance of defining clear Doppler criteria and ensuring sufficient lesion diversity when designing future reliability studies.

We acknowledge that the present work primarily focused on the development of definitions for morphological arterial wall changes, whereas the assessment and definition of haemodynamic abnormalities require further refinement. Arterial stenosis and occlusion have a major impact on the prognosis of patients with TAK; however, haemodynamically significant lesions are typically underpinned by morphological wall changes, and indeed, an arterial stenosis in the absence of ultrasound-visible wall thickening is unlikely to be attributable to TAK. Whether all lesions defined in the present study will ultimately be incorporated into the new OMERACT ultrasound composite score will be determined during the subsequent phases of this project.

Furthermore, the distinction between inflammatory and atherosclerotic wall changes represents an additional challenge in the ultrasound assessment of TAK as reflected by the modest reliability of results in exercise 1 of our study. The consensus-based definition of the 'macaroni sign’ incorporates morphological features that facilitate the discrimination between vasculitis and atherosclerosis, and their correct application, as discussed in the calibration meeting before Exercise 2 led to better reliability results. Refining the operational criteria distinguishing vasculitic from atherosclerotic lesions will also be an important consideration for the development of the future OMERACT ultrasound score.

Further limitations are the small sample size and relatively homogeneous ethnicity, which may limit generalisability. Although a larger and more diverse cohort, both demographically and in terms of lesion prevalence, would have been desirable, patient availability as well as logistical and feasibility constraints inherent to live patient-based reliability exercises imposed practical limits on recruitment. Moreover, we did not apply a convex probe for the evaluation of the abdominal aorta potentially limiting image quality and lesion assessment in some patients; nevertheless, all linear probes were operated with dedicated presets and software adjustments optimised for the assessment of the abdominal aorta. In addition, although the definition of the 'macaroni sign’ was agreed on through a structured Delphi consensus, it remains partly qualitative, which may limit its reproducibility; the upcoming phases of this OMERACT process will focus on incorporating this sign into a quantitative ultrasound composite score, whose reliability will then be formally assessed. Finally, only one patient had clinically active disease at the time of assessment, and our cross-sectional design does not allow us to distinguish ultrasound lesions reflecting acute inflammation from those related to chronic wall remodelling. A prospective, longitudinal study evaluating the sensitivity to change of these elementary lesions on initiation of treatment as well as during a relapse is planned.

Despite these limitations, the application of standardised OMERACT consensus definitions, uniform ultrasound settings, and strict blinding procedures ensured methodological consistency. The inclusion of experienced sonographers from different countries enhanced generalisability.

Looking ahead, these results provide the empirical basis for the next steps of this project, according to the OFISA approach: a Delphi exercise to develop a composite ultrasound score, followed by online and on-site reliability exercises and prospective studies to evaluate and validate the future score.^{7 8} The consistently higher reliabilities observed in the carotid and axillary arteries compared with the other segments support the concept that a core set of easily accessible arterial territories could serve as the basis of the future ultrasound score, maximising reproducibility while still allowing deeper segments (eg, subclavian artery or aorta) to be included to capture the full disease spectrum.¹ A comprehensive ultrasound assessment in TAK requires proficiency in both the evaluation of arterial wall morphology and the assessment of haemodynamic alterations. Although clinicians performing ultrasound in large-vessel vasculitis are expected to have acquired these skills, a structured training curriculum and validated competency-assessment framework remain to be developed.

In conclusion, this study demonstrates the feasibility and reproducibility of the consensus-based ultrasound definitions of elementary lesions in TAK applied by trained experts within a standardised framework. Morphological wall thickening (ie, ‘macaroni sign’) proved to be the most reliable feature, whereas haemodynamic lesions such as stenosis and occlusion require further refinement and validation. These findings contribute to the development of a validated OMERACT ultrasound score for disease monitoring and outcome assessment of TAK patients in clinical trials.

Data availability statement

Data are available upon reasonable request.