Patients
Data of all consecutive children with MRI evaluations of the knees between November 2007 and October 2008 were included in this prospective observational study. Indications for MRI evaluation were clinical active arthritis suspicious for JIA (group 1), or follow-up of children with JIA and clinically inactive disease who had a history of clinically evident arthritis in at least one knee (group 2). Children visited one of the outpatient clinics of two tertiary pediatric rheumatology centers (Emma Children’s Hospital/Academic Medical Center and Jan van Breemen Institute, both in Amsterdam, The Netherlands). At the time of presentation, all children had clinical and laboratory assessments followed by a radiological evaluation of both knees consisting of conventional radiographs and MRI. Children who had had an intra-articular corticosteroid injection in the last 6 months were excluded. The study was performed in accordance with the Declaration of Helsinki and the local medical ethical regulations. Consultation with our institutional review board revealed that informed consent was not necessary, since the MRI approach described is noninvasive.
Children in group 1 met the criteria for JIA defined as clinical arthritis in ≥ 1 joint for > 6 weeks, suggestive for JIA. After a period of 6 months, all newly diagnosed children with JIA were clinically re-evaluated, and reclassified if necessary [
7]. In group 2, clinically inactive disease was defined according to the preliminary criteria for inactive disease in JIA by Wallace et al. [
8].
MRI protocol
To optimize the feasibility of the MRI examination, intravenous contrast injection was omitted and an open-bore MRI scanner was used. Therefore, non-contrast-enhanced MRI of both knees was performed using an open-bore 1-T MRI system (Panorama HFO; Philips Medical Systems, Best, The Netherlands) with a dedicated knee coil. No sedation was used. MRI sequences included sagittal T2-weighted fat-saturated images (TR, 4,000-6,000 ms; TE, 30 ms; slice thickness, 4 mm; field of view, 150 × 92 mm; matrix, 232 × 186; bandwidth, 218.6 Hz/pixel), coronal T2-weighted fat-saturated images (TR, 2,700-4,500 ms; TE, 50 ms; slice thickness, 4 mm; field of view, 150 × 114 mm; matrix, 248 × 195; bandwidth 169.6 Hz/pixel), axial T2-weighted fat-saturated images (TR, 2,900-4,300 ms; TE, 50 ms; slice thickness, 4 mm; field of view, 150 × 114 mm; matrix, 248 × 195; bandwidth, 169.6 Hz/pixel), sagittal T1-weighted images (TR, 450–650 ms; TE, 10 ms; slice thickness, 4 mm; field of view, 150 × 92 mm; matrix, 221 × 234; bandwidth, 178.7 Hz/pixel), and axial turbo spin-echo proton density-weighted images (TR, 3,000-5,000 ms; TE, 30 ms; slice thickness, 4 mm; field of view, 170 × 124 mm; matrix, 340 × 252; bandwidth, 246.3 Hz/pixel). The total scan time was 41 min 28 s for both knees.
Image analysis
MR images were interpreted in consensus by one expert musculoskeletal radiologist (MM; 16 years of experience in musculoskeletal radiology) and a radiology trainee (RH; 3 years of experience in musculoskeletal radiology) masked to clinical history, the duration, extent and severity of the symptoms. A newly composed literature-based grading score was used for assessment [
2,
11‐
16]. Each knee was assessed at the time of MRI interpretation according to the following criteria:
1.
Synovial hypertrophy. The synovium was assessed for signal intensity characteristics and maximal thickness and was evaluated at six sites of the knee joint; patellofemoral, suprapatellar recesses, infrapatellar fat pad, adjacent to the anterior and posterior cruciate ligaments, adjacent to the medial posterior condyle and lateral posterior condyle. Increased synovial thickness of ≥3 mm was considered definitely abnormal [
11]. Synovial thickness was scored based on the maximum thickness in any section as follows: grade 0 if <3 mm, grade 1 if ≥3 mm.
2.
Articular cartilage lesions. As progressive JIA may lead to cartilage destruction, the articular cartilage was assessed for focal destruction (intact, superficial loss and/or thinning, or deep erosions to subchondral bone) [
11]. The cartilage was scored for the presence of lesions at the patella, at the medial and lateral femur, and at the medial and lateral tibia plateau. Cartilage lesions were scored as absent (grade 0) or present (grade 1).
3.
Bone erosions. Bone erosions were defined according to the OMERACT RAMRIS guidelines as sharply marginated bone lesions with low signal intensity in T1-weighted images in at least two planes, with a cortical break present in at least one imaging plane [
2]. The bone was scored for the presence of erosions in the patella, in the medial and lateral femur, and in the medial and lateral tibia plateau (scored as grade 0 if not present, grade 1 if present).
4.
Bone marrow changes, suggestive for bone marrow edema. Bone marrow changes, suggestive for bone marrow edema, were defined as lesions within the trabecular bone, with ill-defined margins and high signal intensity on T2-weighted fat-saturated and low signal intensity on T1-weighted images. The bone marrow was scored for the presence of changes in the patella, in the medial and lateral femur, and in the medial and lateral tibia plateau (scored as grade 0 if not present, grade 1 if present).
5.
Heterogeneity of the infrapatellar fat pad. Infrapatellar fat pad heterogeneity correlates with synovial thickness in early JIA [
11] and was evaluated (scored as: 0, absent; 1, minimal; 2, marked). In addition, fat pad heterogeneity was further classified as being caused by water infiltration or by scar tissue. On MRI, water infiltration of the infrapatellar fat pad is depicted as areas of decreased signal intensity on T1-weighted images and increased signal intensity on T2-weighted fat sat images, whereas infiltration of the infrapatellar fat pad by scar tissue shows decreased signal intensity on both T1-weighted and T2-weighted images [
12].
6.
Joint effusion. The extent of effusion was subjectively graded within the suprapatellar recesses, the lateral recesses, Baker cyst and central recesses (scored as: 0, absent; 1, minimal; 2, marked) [
11,
13‐
15].
7.
Tendinopathy/internal derangement. The medial collateral ligament, lateral collateral ligament, anterior cruciate ligament, posterior cruciate ligament, patellar tendon and the menisci were assessed for pathology (scored as: 0, normal; or 1, abnormal).
8)
Popliteal lymph nodes. The number of popliteal lymph nodes has been reported to correlate with the amount of inflamed synovial volume in rheumatoid arthritis (RA) patients and their presence has also been described in JIA [
11,
14,
16]. Therefore, the number of popliteal lymph nodes was determined.
The primary analyses were based on the first reading. To get an impression of reproducibility, the same readers reviewed the complete data set of MR images 2 years later in an identical manner. The second scoring took place without knowledge of the results of the first.
All children had conventional radiographs taken of both knees within 2 weeks of the MRI examination. For the purpose of this study, the knee radiographs were reviewed and scored in consensus by an experienced pediatric rheumatologist (MvR; 10 years of experience in musculoskeletal radiology) and a radiology trainee (RH; 3 years of experience in musculoskeletal radiology), according to the Dijkstra composite score [
17]. The readers were masked to the clinical status and MRI findings.
Statistical analysis
Descriptive statistics were reported in terms of medians and interquartile ranges for the continuous variables and in terms of absolute frequencies and percentages for the categorical variables. The Mann–Whitney
U test, Fisher exact test, and the chi-square test were used to analyze differences between study groups. All tests assumed a two-tailed probability. A
P value of less than 0.05 indicated a significant difference. The Cohen kappa coefficient was used to quantify the agreement between observer scores [
18]; agreement was classified as follows:
κ <0.40=poor, ≥0.40–0.60 = moderate, >0.60-0.80=substantial and >0.80=good agreement. SPSS Statistics (version 16.0; SPSS, Chicago, IL) was used for all analyses. Statistical analyses were performed in close collaboration with a clinical epidemiologist.