An internationally agreed protocol for the MSUS detection of hip synovitis in children with JIA was developed and tested. The protocol showed good sensitivity and specificity, with good inter-operator reliability. Hip joint involvement in patients with JIA can be frequent (15–50%) and disabling [6, 7, 15]. Insufficient recognition leads to chronic synovitis which can be characterized by pain, stiffness and progressive joint destruction and in the worst-case lead to a requirement for joint replacement surgery [7, 16, 17]. MRI is regarded as the gold standard imaging technique for identification of hip arthritis and the associated findings, due to the depth of the joint, which does not allow the detection of swelling or warmth on clinical examination, as in other joints [18]. Despite this, the clinical assessment of the hip remains a cornerstone. Nistala et al. showed a sensitivity of 25.7% and specificity of 91% for the clinical detection of arthritis when compared to the MRI findings in a cohort of 34 JIA patients with established disease [19]. Ostrowska et al. report a sensitivity of 25% and specificity of 100% of MRI in discriminating between patients with suspected hip disease who later had a confirmed or disapproved JIA diagnosis [20]. Nonetheless, MRI examination requires sedation in the less cooperative children. Furthermore, it is expensive and may not be widely accessible. Therefore, the use of simple and reliable protocols for examination using non-invasive, child-friendly, relatively inexpensive repeatable imaging technique such as MSUS can be of major benefit for clinicians both in clinical care and in research. Indeed, MSUS provides a quick and easily accessible method of differential diagnosis in clinical practice which can be routinely employed. Additional applications include monitoring of arthritis under treatment and the involvement of enthesis and tendons [21]. MSUS is also increasingly being recognised to be equivalent to MRI in some situations, e.g. the evaluation of shoulder and ankle tendon abnormalities, Baker's cyst and wrist ganglion cysts [22] and is superior to clinical examination for the evaluation of synovitis in peripheral joints [23]. Abnormal MSUS findings associated with clinical synovitis was shown by Silva et al. who performed ultrasound of 184 hip joints in patients with JIA with and without hip-related symptoms [24]. The study additionally showed ultrasound changes in 31.5% (29/92) of patients, which was termed subclinical arthritis.

This study showed that BM positivity in the longitudinal view was highly sensitive for the evaluation of synovitis in the anterior recess. In contrast, PD had a very low sensitivity in this region and was therefore not included in the final PIUS-Hip protocol. PD negativity is in line with the hypothesis that the very low intrasynovial blood flow in the depth cannot be sufficiently represented by Doppler technique, even with the newer scanning devices. This limitation of the Doppler modality was also acknowledged in a recent paper from this working group, on the development of a PIUS-Knee protocol. Similarly, PD-positivity had higher sensitivity in more superficial views of the knee joint, including the medial (0.67, 0.59–0.75) and lateral (0.69, 0.60–0.76) parapatellar scans and longitudinal lateral (0.67, 0.60–0.75) views compared to the deeper suprapatellar longitudinal view (0.43, 0.35–0.51) [3].

BM showed to be a very reliable technology to discriminate normal and pathological JIA findings in hip joints. Two clinically affected JIA hips had negative BM findings in all views. These clinical findings were therefore suspected to be due to extra-articular causes, such as muscle strain or soft tissue involvement. Two clinically unaffected JIA hips had a grade I positive semiquantitative B-Mode finding, possibly representing a residual or early ultrasound finding. The semiquantitative joint specific scoring showed an excellent sensitivity and specificity (Supplement 1). Sensitivity of BM grading was marginally higher than the anterior recess size cut-off use in BM, though that had a marginally higher specificity. Therefore, a combination of the semiquantitative BM grading and the quantitative anterior recess measurement would comprise the optimal protocol.

Ultrasound derived anterior recess size has already been deemed a relevant aspect of the hip joint examination [18, 25], whilst a positive correlation with reduced range of movement of the hip joint has also been described [26]. Silva et al. used a cut-off of < 6 mm to define normal, also defined by Frosch et al. and Fedrizzi et al., and found 29% of the measurements were abnormal, though correlations with symptoms or clinical signs were not described [24]. Fedrizzi et al. showed that 50% of their JIA patient sample, also not defined as having clinical pathology of the hip, had an anterior recess size > 6 mm as well as other signs of synovitis in ultrasound including increased echogenicity and distension of the joint capsula [18]. In this study, a cut-off of 7.2 mm was established. Measurements from a previously published independent control cohort of healthy hip joints in children without any arthritis were included in the data analysis in order to avoid the limitation of only including patients with JIA and their hips without clinical arthritis as controls. Indeed, analysis showed no statistically significant differences between the internal (JIA patients, arthritis or healthy hips) and external (no JIA, both hips included) analysis of anterior recess size, indicating that no cases of subclinical JIA arthritis had an impact on the study results. A limitation recognised by the authors is the lack of patients in the youngest age group aged 1–3 years, with the included number increasing with age. This reflects the relative lack of coxarthritis in young patients with JIA.

An anterior transverse view was added to the protocol presented here to visualize the femoral head and to measure the cartilage thickness. In a study published by Spannow and colleagues the cartilage thickness in the knee region was significantly lower in JIA patients in comparison to healthy controls [27]. Although a significant difference in femoral head cartilage thickness between the JIA patients with and without clinical hip arthritis could be seen (Table 2), this difference was not maintained in the age-specific subgroup analyses (Supplement 4). Therefore, the inclusion of the transverse view to measure the cartilage thickness is not mandatory in a routine protocol for evaluating hip synovitis.

Furthermore, this study was able to show the anterior and posterior capsula thickness in hips did not differ between JIA patients with or without hip synovitis, which reflected the findings from the study by Robben et al., which was however performed in patients with transient hip synovitis [14]. Therefore the routine measurement of capsula thickness for the evaluation of hip synovitis is not recommended. Zuber et al. described normal values for the hip joint capsule in 816 hip MSUS examinations in children referred to the rheumatology clinic who subsequently had no musculoskeletal disease diagnosed, showed a relationship to height rather than age, but no difference between the genders. However, the anterior and posterior capsula thickness was not specifically discriminated [12].