Radiologic features in juvenile idiopathic arthritis: A first step in the development of a standardized assessment method

Authors


Abstract

Objective

To describe radiologic features of patients with juvenile idiopathic arthritis (JIA) in a standardized manner, to test the reliability and feasibility of this description, and to correlate these features with clinical signs as a first step in the development of a standardized assessment method.

Methods

The placebo-controlled study of sulfasalazine in patients with oligoarticular, extended oligoarticular, and polyarticular JIA performed by the Dutch Juvenile Idiopathic Arthritis Study Group yielded the data for this study. All trial entry radiographs (clinically involved joints and contralateral joints) were scored (in consensus by a skeletal radiologist and pediatric rheumatologist) for the presence of swelling, osteopenia, joint space narrowing, growth abnormalities, subchondral bone cysts, erosions, and malalignment.

Results

Data on 67 of 69 patients were analyzed. The mean age was 9.1 years (range 2.5–17.6 years), and the median disease duration was 24 months (range 5–176 months). Thirteen percent of the patients were IgM rheumatoid factor (IgM-RF) positive, and 16% were HLA–B27 positive. All 68 clinically evaluated joints were included in the maximum of 19 radiographed joints (or joint groups) per patient. The mean number of radiographed joints per patient was 7 (range 2–15); knees, hands, ankles, and feet were most frequently affected. Fifty-eight patients (87%) had radiologic abnormalities in at least one joint (soft-tissue swelling in 63% of patients, growth disturbances in 48%, joint space narrowing in 28%, and erosions in 15%). In total, half of the radiographs of the clinically involved joints showed radiologic abnormalities, including two-thirds of the radiographs of the clinically affected hands and knees. Univariate analysis revealed a good correlation between the overall articular (clinical) severity and the presence of radiologic abnormalities (odds ratio [OR] 1.38, P < 0.0001). Multivariate analysis showed increased ORs for the presence of radiologic abnormalities and IgM-RF positivity (OR 4.6, P = 0.005) or HLA–B27 positivity (OR 3.0, P = 0.004). In general, reproducibility of the radiologic scoring method was good (mean kappa coefficient of 0.74 [range 0.40–0.86]), although there were scoring discrepancies for swelling, osteopenia, and growth disturbances. The scoring took 10–20 minutes per patient.

Conclusion

Our model of describing and scoring radiologic abnormalities of radiographed joints in JIA was feasible, mostly reproducible, correlated well with the overall articular severity score, and added substantial new information not available on clinical examination.

Juvenile idiopathic arthritis (JIA; previously called juvenile chronic arthritis [JCA]) (1) can lead to destructive lesions of joint cartilage and periarticular bone. As in adult rheumatoid arthritis (RA), radiographs are deemed important to document this damage and are used widely by clinicians to assess disease severity and progression. The extent of reported radiologic damage varies with the type of JIA patients studied and the definition of radiologic damage. In recent long-term outcome studies of patients with oligoarticular-onset and polyarticular-onset JIA, joint damage in up to 70% of patients was observed (2, 3).

Compared with RA, the radiologic manifestations of JIA differ with regard to 1) the number and distribution of involved joints (e.g., in oligoarticular JIA, fewer than 5 joints are involved, usually asymmetrically; the most frequently involved joints are the knee, ankle, elbow, and hand); 2) the manifestation of joint space narrowing (articular cartilage is generally thicker in children than in adults, and radiographic narrowing of a joint space is the reflection of cartilage loss [4]); 3) the development of growth disturbances in JIA (inflammation occurs in a developing joint); 4) joint ankylosis (joint ankylosis occurs more promptly in children than in adults, particularly in the carpal and tarsal joints and in the cervical spine) (5); and 5) joint erosions, which develop later in the disease course in JIA (6–10). Thus, although radiologic scoring systems are most advanced in RA, any scoring system in JIA modeled on this experience must take these unique features of JIA into account.

Petterson and Rydholm made the first attempt to develop an objective radiologic scoring system for joint abnormalities of the large joints in JIA (11, 12). Later, carpal length measured by plain-film radiography (specifically, the ratio of the carpal length to the length of the second metacarpal) was introduced as a radiologic assessment technique for cartilage integrity (13, 14). Both techniques have their limitations in use, since they are not applicable to all of the possibly involved joints in JIA.

A series of radiographs comprise the simplest and cheapest permanent record of the cumulative joint damage caused by the disease. Other methods of imaging, such as magnetic resonance imaging and ultrasonography, may offer unique, useful features (15, 16) but are unlikely to replace plain radiography as the standard for some time to come (17, 18). Because JIA therapy is moving toward early treatment with potentially toxic second-line antirheumatic drugs, there is a growing need for a clear and reproducible radiologic assessment standard, both to select and to evaluate patients. The aim of our study was to make a first step in the development of a standardized assessment method applicable to radiographs of patients with JIA, and to assess the reliability, feasibility, and measurement properties of this method. The sensitivity to change and other methodologic issues of evaluation will be addressed in future studies.

The issues that we aimed to explore in this radiologic study were as follows: 1) assessment of the presence or absence of a comprehensive spectrum of radiologic features among JIA patients who took part in the sulfasalazine (SSZ) placebo-controlled study performed in the Netherlands (19), 2) testing of the reliability and feasibility of this assessment, and 3) evaluation of the correlation between these radiologic features and clinical signs.

PATIENTS AND METHODS

Cohort.

The placebo-controlled SSZ study performed by the Dutch Juvenile Idiopathic Arthritis Study Group (19) yielded the data for the present study. Data on patients who had a complete clinical and laboratory assessment with a complete set of radiographs at study entry were included. All patients included in the original study fulfilled the European League Against Rheumatism criteria (20) for oligoarticular- or polyarticular-onset JCA and were between the ages of 2 years and 18 years, with onset of arthritis before the age of 16 years. For the purpose of the present study, patients were retrospectively reclassified according to the JIA subtypes.

Clinical data.

In every patient, 68 joints were scored for swelling (range 0–3), pain on motion and/or tenderness (range 0–3), and limitation of motion (LOM) (range 0–4) (19). The overall articular severity score was defined as the sum of all scores for swelling, tenderness/pain, and LOM (range 0–10) (21). In this trial, “clinical arthritis” was defined as the presence of swelling or LOM, in addition to either pain upon movement or tenderness. In the present study, joints with either swelling, pain, or LOM (i.e., overall articular severity score ≥1) are described as joints with “clinical signs of disease” and include joints with clinical arthritis. For this radiologic study, we used all data on the radiographed joints/joint groups (maximum of 19 radiographed joints/joint groups per patient).

Radiologic data.

Collection of radiographs.

At entry into the SSZ/placebo trial, conventional film-screen radiographs of all affected joints (either tender, painful, swollen, or limited in motion as judged by the treating physician) and the contralateral joints were obtained; these radiographs were used in the present study. Information about the duration of disease in the radiographed joint was not systematically collected. Although followup films were made in the trial, this report is limited to the baseline radiographs.

Reading method.

After completion of the SSZ/placebo trial, the radiographs were read in chronologic order in a single session by a skeletal radiologist (PFD) and a pediatric rheumatologist (MAJvR) in consensus. The readers were unaware of the subtype of JIA and the clinical condition of the patient. All primary analyses were based on this first reading. To get an impression of reproducibility, these same readers reviewed a convenience sample (from the Leiden and Amsterdam centers) of these same radiographs 4 years later, with the same objective and in an identical manner. The second scoring took place without knowledge of the results of the first scoring. Finally, the readers revised all radiograph findings that had discrepant scores between the first and second session, but now with the knowledge of both scores.

Joints scored.

The following 19 joints/joint groups were evaluated by radiography: cervical spine (1 joint), 2 mandibles, 2 shoulders, 2 elbows, 2 hands (each hand comprises a joint group that includes all finger, metacarpal, and wrist joints), 2 sacroiliac joints, 2 hips, 2 knees, 2 ankles, and 2 feet (each foot comprises a joint group that includes all tarsal, metatarsal, and toe joints). In this report, the term “joint” will be used for both a large single joint or a group of smaller joints (e.g., hands).

Features scored.

A scoring list applicable to these 19 joints/joint groups was composed. The following features (collectively called radiologic abnormality) were scored: soft-tissue swelling, osteopenia, joint space narrowing, enlargement or other growth disturbances, subchondral bone cysts, erosions, and joint position or alignment. A combination of radiologic joint scores for RA served as inspiration for the scoring of soft-tissue swelling, grading of erosions, and joint alignment (22, 23).

High-intensity light was used to assess soft-tissue swelling, and this was scored as present if a reflection of soft-tissue swelling was found around a joint. Subchondral osteoporosis was defined as present when a localized decrease of bone density was noticed around a joint. Growth abnormalities were analyzed with regard to the shape, development, and maturation of the bone (24, 25). Overgrowth of the epiphyses was assessed by measurement of the epiphyses and comparison with the contralateral side (26). Special attention was paid to acceleration of epiphyseal maturation, premature fusion of the epiphysis (eventually resulting in an abnormally short bone), tapering of the juxta-epiphyseal parts of the shaft of the bone, deformity of the joint due to asymmetric growth, and modeling abnormalities (e.g., “ballooned” epiphysis, squared patella, abnormal shape of carpal bones) (10). Joint morphology was assessed with reference to age- and sex-adjusted standards (27). If any of the above-mentioned growth abnormalities occurred, the joint/joint group was scored positive for growth abnormalities.

Subchondral bone cysts were defined as localized areas of bone destruction and scored positive when present around a joint. Erosions were defined as a discrete interruption of the cortical surface of the bone. Erosions were scored as absent or present, and graded according to the amount of destruction of the joint surface (DJS): DJS <25%, DJS 26–50%, DJS 51–75%, and DJS >75% (22). In the present study, results of grading of the DJS are not reported. In the assessment of joint alignment, abnormal joint position was defined as flexion deformity (e.g., proximal interphalangeal joints, hands), subluxation (e.g., boutonnière deformity), or dislocation (e.g., ulnar and radial deviation). One joint/joint group could have a positive score for more than one feature, but each feature was counted only once (e.g., one joint/joint group could show more than one erosion, but the erosion score remained 1).

Index joint.

An index joint can be described as a joint (or combination of joints) that represents the damage in all of the joints (i.e., the presence and extent of damage in this joint is highly correlated with the damage in the other joints of a patient). We investigated whether hand, foot, or knee joints could be used as an index joint in JIA.

Laboratory data.

The laboratory data at study entry included the erythrocyte sedimentation rate, as well as positivity for antinuclear antibodies (ANA), IgM and IgA rheumatoid factor (IgM-RF and IgA-RF, respectively), and HLA–B27.

Statistical analysis.

The presence of radiologic abnormalities was summarized both at the level of the various joints and at the level of the patient. Standard and logistic regression models that included a random patient effect analyzed the association between different types of radiologic abnormalities, and between the presence/absence of radiologic abnormalities in the various joints and the characteristics of the patient. P values of less than or equal to 0.05 were considered significant. Cohen's kappa coefficient quantified the agreement between observer scores (28).

RESULTS

Patient characteristics.

The placebo-controlled SSZ study included 69 patients with JIA. For the present study, 2 male patients were excluded: 1 was reclassified as having systemic JIA, and 1 had missing radiographs. Onset of JIA was oligoarticular in 37 patients and polyarticular in 30 patients (Table 1). The underlying diseases, using the JIA criteria (1), were as follows: persistent oligoarthritis (19 patients [28%]), extended oligoarthritis (8 [12%]), IgM-RF–negative polyarthritis (19 [28%]), IgM-RF–positive polyarthritis (9 [13%]), enthesitis-related arthritis (7 [10%]), arthritis and psoriasis (1 [1%]), and other arthritis (4 [6%]). In each patient, a mean of 12.8 joints (range 1–68) showed clinical signs of disease, and of these joints, a mean of 5.0 (range 1–42) fulfilled the definition of clinical arthritis (Table 2).

Table 1. Characteristics at study entry of 67 patients with juvenile idiopathic arthritis (JIA) who participated in the placebo-controlled trial of sulfasalazine and whose radiographs were used for radiologic scoring evaluation*
CharacteristicValue
  • *

    Except where otherwise indicated, values are the no. (%) of patients. IQR = interquartile range; DMARD = disease-modifying antirheumatic drug.

  • Clinical arthritis defined as the presence of swelling or limitation of motion, plus pain upon movement or tenderness; total no. includes 12 patients with oligoarticular-onset JIA and a polyarticular disease course, and 1 patient with polyarticular-onset JIA with <5 joints with clinical arthritis at study entry.

Age, mean (SD) years [range]9.1 (4.1) [2.5–17.6]
Female46 (69)
Disease onset before age 6 years35 (52)
Disease onset at age 6–10 years16 (24)
Disease onset beyond age 10 years16 (24)
Disease duration, median (IQR) months [range]24 (10–40) [5–176]
Polyarticular-onset type JIA30 (45)
Oligoarticular-onset type JIA37 (55)
>4 joints with clinical arthritis at study entry41 (61)
Antinuclear antibodies present33 (49)
IgM rheumatoid factor present9 (13)
HLA–B27 positive11 (16)
Local corticosteroid use ever30 (45)
DMARD medication use ever5 (7)
Systemic corticosteroids ever2 (3)
Table 2. Per patient distribution of clinical signs and radiologic abnormalities in the radiographed joints*
VariableNumber of joints with clinical signs or radiologic abnormalityNumber of radiographed joints
MeanRangeMeanRange
  • *

    The term “joints” is used to describe either a single joint (e.g., metacarpal joint or knee joint) or a group of smaller joints (e.g., cervical spine, wrist).

  • From among a total of 68 clinically evaluated joints/joint groups.

  • Defined as joints with either swelling, pain, or limitation of motion, but not satisfying the definition of clinical arthritis.

  • §

    Defined as joints with swelling or limitation of motion, plus pain upon movement or tenderness.

  • From among a maximum of 19 radiographed joints/joint groups per patient.

Clinical sign
 Swelling8.40–443.70–10
 Pain6.20–422.30–14
 Limitation of motion7.00–613.70–17
 Clinical signs of disease12.81–685.91–17
 Clinical arthritis§5.01–422.50–14
Radiologic abnormality
 All scored features2.80–117.02–15
 All scored features excluding soft-tissue swelling2.30–97.02–15

Radiologically evaluated joints.

Radiographs had been obtained for 94% of the swollen joints, 81% of the painful joints, 83% of the joints with LOM, and 85% of the joints with clinical arthritis. In total, 471 joints/joint groups were radiographed. The mean number of radiographed joints/joint groups per patient was 7 (range 2–15), and of these radiographed joints/joint groups, a mean of 5.9 (range 1–17) had clinical signs of disease in at least one of the joints of the joint group, and a mean of 2.5 (range 0–14) had clinical arthritis. Per patient, a mean of 2.8 (range 0–11) radiographed joints/joint groups showed radiologic abnormalities (Table 2).

Radiologic findings.

Fifty-eight patients (87%) had radiologic abnormalities in at least one of their radiographed joints. The most frequent abnormality was soft-tissue swelling (63% of the patients), followed by growth disturbances (48%), joint space narrowing (28%), and erosions (15%) (Figure 1). With regard to the distribution of abnormalities among the radiographed joints, soft-tissue swelling was seen in 19% of the radiographed joints, followed by growth disturbances (11%), joint space narrowing (8%), and erosions (6%) (Figure 1). The most diverse radiologic signs of disease were noted in the hand, as well as in the knee, ankle, and foot. Erosions were detected in one knee, while all other erosions were scored in the hand or the foot. The correlations between different radiologic features were modest (range r = 0–0.30), except for the relationship between osteopenia and abnormal joint position (r = 0.51).

Figure 1.

Proportion of patients with juvenile idiopathic arthritis showing a specific radiologic abnormality in at least one of their radiographed joints/joint groups (solid bar), and proportion of joints/joint groups showing the specific radiologic abnormality (gray shaded bar). Note that one patient or one joint/joint group can have more than one abnormality.

An index joint could not be identified. Of the 67 patients, 49 had either hand or foot radiographs. Of the 41 patients with hand radiographs, 22 showed radiologic abnormalities in the hands, together with radiologic abnormalities in other joint groups; of the remaining 19 patients with hand radiographs, 17 had no radiologic abnormalities in the hands despite having abnormalities in other joints, and 2 patients had radiologic abnormalities in only the hands. Exclusion of the radiologic sign of soft-tissue swelling from the analysis did not substantially alter the findings. Similarly, neither the foot, the knee, nor a combination of hand, foot, and knee qualified to serve as an index of radiologic damage in the other joints (data not shown).

Association of radiologic findings and clinical signs.

In 82% of radiologically swollen joints, clinical swelling coincided with the radiologic swelling in at least one of the joints of the joint group. In 92% of joints with joint space narrowing, clinical signs of disease were present in at least one of the joints of the joint group. Moreover, clinical signs of disease were present in 84% of joints with growth abnormalities and 65% of joints showing erosions. All cases of radiologically abnormal joint position showed clinical signs of disease.

As noted before, 85% of joints with clinical arthritis were radiographed, and half of these radiographed joints showed abnormalities (Tables 3 and 4). Radiographs of the clinically affected hands and knees most frequently showed detectable radiologic abnormalities (66% and 64%, respectively). Of note, a relatively high proportion of hand and foot radiographs (36% and 39%, respectively) showed abnormalities without distinct clinical symptoms, while in the knee, ankle, and elbow, this proportion was lower (23%, 10%, and 6%, respectively).

Table 3. Clinical and radiologic abnormalities in 471 radiographed joints (joint groups) in 67 patients with juvenile idiopathic arthritis
Joint site*Total radiographed joint groups (n = 471)Joint groups with clinical arthritis (n = 186)Joint groups with clinical signs of disease (but no clinical arthritis) (n = 130)Joint groups without clinical signs of disease (n = 155)
Total no. radiographs% abnormalNo. radiographs% abnormalNo. radiographs% abnormalNo. radiographs% abnormal
  • *

    Total number of sites: maximum 19 joints/joint groups per patient (i.e., cervical spine, knee left, knee right, etc.).

  • Except for missing data on one right mandible, all joints (joint groups) were radiographed at both sides.

Knee10447426431483123
Ankle8427353120452910
Foot7036204519213139
Hand8657556620451136
Elbow4217183370176
Shoulder30350138120
Hip28186171040120
Cervical spine1863080714
Sacroiliac joint667210000450
Mandible30002010
Total47135186501303215520
Table 4. Distribution of radiologic abnormalities over the radiographed joints (joint groups) with “clinical arthritis” in at least one of the joints of the joint group*
Joint siteNumber of radiographed joint groups with clinical arthritisNumber of radiographs with abnormalitiesSpecification of radiologic abnormalities, joint counts
SwellingOsteopeniaJoint space narrowingGrowth abnormalBone cystsErosionsAbnormal joint position
  • *

    No abnormalities were found in the radiographs of the cervical spine, shoulder, and mandible (10 films). NA = not applicable.

  • Values in parentheses are the number of joint groups with >1 radiologic sign.

Knee4227 (9)2035111
Ankle3511 (2)9112
Foot209 (6)21345
Hand5536 (33)2016201321014
Elbow186 (0)24
Hip61 (1)NA111
Sacroiliac joint22 (0)NA2

The correlation between clinical symptoms and radiologic findings was analyzed further. When the overall articular severity score of the radiographed joint was related to the presence of radiologic abnormalities of these joints, the results showed a significant increase in the probability of radiologic signs with each increase in the overall articular severity score, using the logistic regression model. Because the overall articular severity score includes a score for swelling, we excluded the radiologic sign of swelling for this analysis. The odds ratio (OR) for the overall severity score was 1.4 (P < 0.0001). The relationship between the overall severity score and the probability of radiologic signs is given in Figure 2. A total of 155 radiographed joints had an overall severity score of 0 (no clinical signs of disease), but 20% of these showed radiologic abnormalities (marginal probability of 0.20) (Figure 2).

Figure 2.

Correlation between clinical symptoms and presence of radiologic signs. The overall severity score of a joint is defined as the sum of the severity ratings of swelling (range 0–3), pain/tenderness (range 0–3), and limitation of motion (range 0–4) (21). In the calculation of the marginal probability of radiologic signs, the radiologic sign “soft-tissue swelling” was not included. The thick solid line represents the computed marginal probability of radiologic signs given the overall severity score according to the logistic model. The dotted lines with shaded areas represent the 95% confidence intervals. Solid circles denote the observed proportion of patients with radiologic signs, and the adjacent values are the number of radiographed joints/joint groups from which the proportions are calculated.

In explanatory univariate analysis, ORs for the presence of radiologic abnormalities were significantly increased in relation to the overall articular severity score, onset of arthritis beyond 10 years of age, IgM-RF positivity, or IgA-RF positivity (Table 5). Multivariate analysis showed a significant increase of the OR for the presence of radiologic abnormalities (excluding swelling) and IgM-RF positivity (OR 4.6, 95% confidence interval [95% CI] 1.44–14.5, P = 0.005) as well as HLA–B27 positivity (OR 3.0, 95% CI 1.32–6.84, P = 0.004). Inclusion of radiologically described swelling in the score did not significantly change the results of this multivariate analysis (data not shown).

Table 5. Odds ratio (OR) for the presence of at least one radiologic abnormality (excluding radiologically described swelling or swelling and osteopenia) in relation to clinical signs in 67 patients with juvenile idiopathic arthritis (JIA) by univariate analysis*
VariableRadiologic abnormalities
Excluding soft-tissue swellingExcluding soft-tissue swelling and osteopenia
OR (95% CI)POR (95% CI)P
  • *

    95% CI = 95% confidence interval; RF = rheumatoid factor; ANA = antinuclear antibodies; ESR = erythrocyte sedimentation rate; NS = not significant (significance level P ≤ 0.05).

Overall articular severity score1.4 (1.2–1.6)<0.00011.4 (1.2–1.5)<0.0001
IgM-RF positive2.4 (1.4–4.0)0.0012.5 (1.5–4.1)0.001
IgA-RF positive2.3 (1.2–4.6)0.022.4 (1.2–4.7)0.01
HLA–B27 positive2.4 (1.4–4.2)0.0012.3 (1.4–4.0)0.002
ANA-positive0.5 (0.3–0.7)0.0010.5 (0.3–0.7)0.001
ESR
 <20 mm/hour1.0 1.0
 >20 mm/hour2.1 (1.1–4.2)0.032.3 (1.1–4.9)0.02
Disease onset age
 <6 years1.0 1.0
 6–10 years1.6 (0.9–2.7)NS1.5 (0.9–2.6)NS
 >10 years2.8 (1.7–4.5)0.00012.8 (1.7–4.7)<0.001
Disease duration
 <2 years1.0 1.0
 2–5 years1.8 (0.9–3.8)NS1.5 (0.7–3.4)NS
 >5 years0.6 (0.2–1.7)NS0.7 (0.2–2.0)NS
Onset type JIA
 Oligoarticular1.0 (0.7–1.5)NS1.0 (0.7–1.5)NS
 Polyarticular1.1 (0.7–1.7)NS1.1 (0.7–1.8)NS

Reproducibility.

To evaluate reproducibility of the readings, we reviewed 240 radiographed joints (120 at baseline and 120 at the 24-week followup) of 15 patients. The overall agreement on the presence of any abnormality showed a Cohen's kappa coefficient with a mean of 0.74. Kappa values of the individual features scored ranged between 0.40 and 0.86, as shown in Table 6. Scores of osteopenia, growth abnormalities, and swelling were the most often discrepant. On review, evaluation of osteopenia and swelling appeared highly dependent on the quality of the radiographs. In the scoring of growth abnormalities, assessing advanced bone maturation or recognition of developmental growth disturbance (e.g., bone shape) appeared especially challenging.

Table 6. Reproducibility of the radiologic scoring method*
Scored featureTotal number of evaluable radiographsAbsolute agreement, %Cohen's kappa
  • *

    Two observers scored radiographs in consensus in one session; the second reading was 4 years later.

  • Total number of evaluable radiographs in which the feature was scored at least once in the two sessions.

  • See ref. 28 for description of the kappa coefficient of Cohen.

No signs of disease167900.74
Soft-tissue swelling126870.66
Subchondral osteopenia123850.40
Joint space narrowing126950.86
Growth abnormalities125860.61
Subchondral bone cysts117970.80
Erosions128940.79
Abnormal joint position121950.76

Feasibility.

Scoring of all the radiographs was time-consuming and required specific expertise on normal bone development and variants of bone development in health and disease. The current extensive consensus scoring took 2 observers ∼10–20 minutes per assessment of each patient.

DISCUSSION

This study is the first step in the development of a standardized scoring system of joint radiographs in JIA. Although, in our study, radiologic damage was related to clinical disease, the data suggest that radiographs often yield important information not available on clinical assessment, and that the scoring of radiologic abnormalities may become feasible in future trials of JIA. Erosions, joint space narrowing, and malalignment were readily and reproducibly identified, but soft-tissue swelling, osteopenia, and growth disturbance caused more difficulty. Total scoring time was acceptable for a trial setting. Further selection of features to score and joints to examine will depend on analysis of responsiveness, which is now in progress.

Joint space narrowing and growth disturbances are considered key radiologic manifestations of JIA. These manifestations have been extensively described in several studies and reviews (6–8, 29, 30), with regard to the cervical spine (31), hips (32), knees (11, 33, 34), carpal length (14), and distal interphalangeal joints (35). A thorough knowledge of these manifestations is necessary to recognize the abnormalities on the radiographs, as is knowledge of the normal development and growth of the different joints in children and adolescents. We were able to score joint space narrowing reliably, but assessment of growth abnormalities was more challenging. Analysis of the radiographs that were scored discrepantly showed that descriptions and definitions of growth disturbance need further refinement. Based on our experience, an atlas with specific definitions may have to be developed to improve scoring reproducibility of this key feature in JIA.

For the clinician, soft-tissue swelling and osteopenia on the radiograph are important early signs of joint involvement in JIA (10). However, both features are best considered signs of disease activity, rather than an indication of damage (2, 7, 8). Although swelling was the most frequently scored radiologic sign of disease (19% of radiographed joints), scores were only moderately reliable because the quality of the radiographs varied (κ = 0.66). Similarly, scores of subchondral osteoporosis and general osteopenia, which were shown to be highly dependent on radiologic technique, differed appreciably between sets of films and were subtle and difficult to reproduce (κ = 0.40). Thus, both swelling and osteopenia may be of limited use in a final scoring system. Localized dual x-ray absorptiometry may be a promising alternative for the study of osteopenia (36).

We noticed bone cysts in the joints in 10% of our studied patients, and this feature showed good reproducibility (κ = 0.80). In our study, only 15% of the patients showed erosions on the radiographs; in all but one radiograph, the erosions were found in the hands and feet. None of the erosions showed a destruction of joint surface of >50%. It is well known that erosions in children develop later in the course of the disease as compared with that observed in RA, and their development depends on the type of JIA (37). In our study population, the median duration of disease was only 2 years.

Apart from the features to score, the other main issue is which joints to assess. In our study (as in other studies), knees, ankles, hands, and feet were most frequently affected (10, 29, 33, 34). In addition, hands and feet showed very diverse radiologic features. Ideally, the abnormalities in one index joint group would be a summary of the abnormalities present in other joint groups. In adult RA, hands and forefeet are widely accepted as such an index. In JIA, some studies have used carpal length as the index, but these included only patients with polyarthritis and those with systemic JIA who had clinical involvement of the wrist (13, 38). In our study, only 58% of the patients had clinical wrist involvement (data not shown), and the abnormalities found in hand radiographs did not correlate well with abnormalities found in other joint groups. The same was true for the other frequently affected joints. Pending further information, we suggest studies should radiograph (and follow up) all clinically affected joints, including, as a minimum, the hands, feet, and knees.

Although univariate analysis showed a good correlation between the overall articular (clinical) severity and the presence of radiologic abnormalities, this relationship was largely unpredictable in specific joints. Many abnormalities were found in clinically uninvolved joints, especially in the hands and feet. This may be partially explained by incomplete information about the history of clinical involvement of these joints. However, we are confident that radiographs add substantial information not obtainable by physical examination.

Some interesting relationships between laboratory features and radiologic abnormalities were seen. In univariate analysis, ANA positivity was associated with the absence of radiologic abnormalities, but this finding was not confirmed in multivariate analysis. In contrast, IgM-RF and HLA–B27 positivity were both independently associated with the presence of radiologic abnormalities. IgM-RF–positive patients with JIA are known to have a disease course similar to that of adult RA, and HLA–B27–positive patients (boys with disease onset beyond 8 years of age and with enthesitis and arthritis) are thought to have a disease course similar to that in adults with ankylosing spondylitis. One other study has also indicated that HLA–B27 positivity in JIA patients relates to a more severe course of JIA (39).

We again note that, in this first step, we did not evaluate sensitivity to change or the relationship to future joint impairment. The long-term functional outcome in the joint is said to be related to joint space narrowing, growth disturbance, and joint deformation by erosions (37). However, it is not clear what type of abnormalities were analyzed to come to this conclusion. For a further development of our scoring method, a long-term followup study is in progress to solve the issues of sensitivity to change and prognostication.

In conclusion, this study shows that scoring radiographs in patients with JIA is feasible and adds information about joint involvement. However, both a thorough knowledge of the radiologic manifestations of JIA and knowledge of the normal development and growth of the different joints is necessary to reliably detect abnormalities on the radiographs. The choices to be made regarding methodologic issues will depend on the purpose for which imaging is done, whether it is to classify, to prognosticate, or to measure change in the joints over time, and might be different for each purpose.

Ancillary