Antinuclear antibody–positive patients should be grouped as a separate category in the classification of juvenile idiopathic arthritis

Authors


Abstract

Objective

We undertook this study to test the hypothesis that in the International League of Associations for Rheumatology (ILAR) classification of juvenile idiopathic arthritis (JIA), patients with similar characteristics can be classified into different categories. We sought to investigate whether antinuclear antibody (ANA)–positive patients having disease in the ILAR categories of oligoarthritis, rheumatoid factor–negative polyarthritis, psoriatic arthritis, and undifferentiated arthritis share homogeneous features and to compare these features with those of ANA-negative patients having the same categories of disease.

Methods

We identified JIA patients who had been followed up during a 22-year period. ANA positivity was defined as ≥2 positive results at a titer of ≥1:160. Demographic and clinical features were recorded retrospectively and compared between ANA-positive and ANA-negative patients.

Results

Of a total of 971 patients, 711 were ANA positive, 149 were ANA negative, and 111 had an indeterminate ANA status. Patients with indeterminate ANA status were excluded. ANA-positive patients in the different ILAR categories were similar in terms of age at disease presentation, female-to-male ratio, and frequency of asymmetric arthritis and iridocyclitis. Compared with ANA-positive patients, the ANA-negative group was older at disease presentation and had a lower prevalence of females, a lower frequency of iridocyclitis and asymmetric arthritis, a greater number of affected joints over time, and a different pattern of arthritis. The close relationship between the presence of ANAs and younger age at disease presentation, female predominance, asymmetric arthritis, development of iridocyclitis, lower number of affected joints over time, and lack of hip involvement was also confirmed by multivariate and multiple correspondence analysis.

Conclusion

Our findings substantiate the hypothesis that ANA-positive patients classified into different JIA categories by current ILAR criteria constitute a homogeneous patient population.

Juvenile idiopathic arthritis (JIA) is an umbrella term that encompasses all forms of arthritis that begin before the age of 16 years, persist for more than 6 weeks, and are of unknown cause (1). Over the years, several classification systems have been developed, none of which has been universally embraced (2–4). The most recent one has been proposed by the International League of Associations for Rheumatology (ILAR) (4) and is primarily aimed at identifying homogeneous disease groups to facilitate research on etiopathogenesis and epidemiology, outcome studies, and treatment trials. The ILAR classification recognizes the following 7 disease categories on the basis of features present in the first 6 months of illness: systemic arthritis, rheumatoid factor (RF)–positive polyarthritis, RF-negative polyarthritis, oligoarthritis, psoriatic arthritis, enthesitis-related arthritis, and undifferentiated arthritis. The oligoarthritis category is divided into 2 subsets: persistent oligoarthritis, in which arthritis remains confined to ≤4 joints throughout the whole disease course, and extended oligoarthritis, in which arthritis extends to >4 joints after the first 6 months of illness.

The ILAR classification has the merit of having resolved the previous disparity in terminology between Europe and North America. However, it has the limitation of not being data-driven, but rather based on expert consensus. Nonetheless, it has been recommended that the current classification system be viewed as “a work in progress,” and pediatric rheumatologists have been urged to participate in the process by making their opinions known and by testing the proposed criteria in their patient series (5).

In recent years, several investigators have evaluated the ILAR criteria and offered numerous suggestions for revision (6–14). In 2003, one of us hypothesized, based on the data in the literature, that a seemingly homogeneous patient group characterized by the presence of antinuclear antibodies (ANAs), early onset of disease, strong predominance of females, prevalence of asymmetric arthritis, and risk for iridocyclitis was classified into 3 different JIA categories (oligoarthritis, RF-negative polyarthritis, and psoriatic arthritis) and that the number of affected joints in the first 6 months of disease as well as the presence of psoriasis did not represent useful criteria to identify homogeneous disease entities in JIA (15). This hypothesis was supported by a subsequent study, in which we showed that ANA-positive patients grouped in the categories of persistent oligoarthritis, extended oligoarthritis, and RF-negative polyarthritis were very similar in terms of age at disease presentation, female-to-male ratio, and frequency of asymmetric arthritis and iridocyclitis (16).

In the present study, we investigated further the role of ANA in identifying a potentially homogeneous disease subset by examining a much larger patient population. Furthermore, we extended the analysis to the ILAR categories of psoriatic arthritis and undifferentiated arthritis to confirm the hypothesis that the ANA-positive patients included in these categories share the same homogeneous features documented in ANA-positive patients with oligoarthritis and RF-negative polyarthritis.

PATIENTS AND METHODS

Patient selection.

This was a retrospective cohort study involving patients who were followed up at the Fondazione Istituto di Ricovero e Cura a Carattere Scientifico (IRCCS) Policlinico San Matteo of Pavia or at the IRCCS G. Gaslini of Genoa between January 1987 and December 2008. Inclusion criteria were a diagnosis of JIA based on the 2001 revised ILAR criteria (4) and a disease category of persistent oligoarthritis, extended oligoarthritis, RF-negative polyarthritis, psoriatic arthritis, or undifferentiated arthritis. Patients were excluded if they met the ILAR criteria for systemic arthritis, enthesitis-related arthritis, or RF-positive polyarthritis because these ILAR categories represent well-defined and separate disease entities.

Included and excluded patients had their ILAR category classified independently by 2 investigators (SO and AMag) based on the review of clinical charts. Discordances were resolved, after discussion, by consensus on the final diagnosis among the 2 investigators and the principal investigator of the study (AR). All patients seen before the publication of ILAR criteria for JIA were reclassified using such criteria. For the sake of uniformity, patients with oligoarthritis were subclassified in the persistent or extended subtype on the basis of the course of joint disease in the first 24 months after disease presentation. The study was approved by the Ethics Committee of the IRCCS G. Gaslini of Genoa.

Classification of ANA status.

Patients had their ANA status defined if they had at least 2 ANA determinations made at least 3 months apart during followup. Our gathering of this information was facilitated by our policy of repeating ANA determinations several times, generally every 6–12 months, in patients with the JIA categories under study. Patients were defined as being ANA positive if they had at least 2 positive results on indirect immunofluorescence at a titer of ≥1:160 and as ANA-negative if they had negative results in all determinations made during the entire followup period. The substrate used for the ANA determinations was rat liver during the first 2 years of the study and HEp-2 cells thereafter. ANA positivity or negativity was confirmed in at least 2 HEp-2 cell assays in all patients for whom the initial assays were done using rat liver as substrate. Patients who had only 1 ANA determination available or who had ≥2 ANA determinations but did not meet the criteria for ANA positivity or negativity were classified as having an “indeterminate” ANA status.

Clinical assessments.

Patients were identified through existing databases and/or clinic files. The medical charts were reviewed for the following information: sex, age at disease presentation, duration from disease presentation until last followup visit, occurrence of iridocyclitis, presence of HLA–B27, asymmetry of arthritis at disease presentation and 6 months after presentation, number and type of joints involved in the first 6, 12, and 24 months after disease presentation, and development of radiographic joint lesions (defined as the presence of joint space narrowing and/or bone erosions in at least 1 joint).

The date of disease presentation was defined as the date when the first symptoms of arthritis were noted, obtained by history as recorded in medical charts. We routinely perform and record on a standardized form a detailed joint assessment in each patient at each clinic visit, which facilitated the gathering of information about joint involvement at specific time intervals. All joint assessments were performed by 3 investigators (AR, SM-M, and Stefania Viola, Genoa, Italy), who used the same methodology. Arthritis was defined as symmetric if >50% of the joints involved during the first 6 months of disease were symmetric pairs. This definition of symmetry was adapted from that used in adults with rheumatoid and psoriatic arthritis, which requires that at least 50% of involved joints be symmetric pairs (17). Eye examinations were performed every 3–4 months in ANA-positive patients and every 6–12 months in ANA-negative patients. Diagnosis of uveitis was confirmed by an ophthalmologist in all patients who developed this complication. All eye examinations performed in the study centers were performed by the same ophthalmologist (Dr. A. M. Broglia in Pavia and Dr. R. De Marco in Genoa) throughout the entire study period.

Clinical data for each patient were collected by investigators using a standardized form developed specifically for this study. Data were then entered into an electronic database.

Statistical analysis.

Descriptive statistics were reported as medians and interquartile ranges for continuous variables and as absolute frequencies and percentages for categorical variables. Comparisons of quantitative variables between 2 groups were made by means Mann-Whitney U test, whereas comparisons of quantitative variables between more than 2 groups were made by nonparametric analysis of variance (Kruskal-Wallis test). Dunn's test was chosen as an a posteriori test to assess the statistical significance of differences between pairs of groups. Categorical data were compared by chi-square test, or by Fisher's exact test in case of expected frequencies <5. Bonferroni adjustment was applied as a correction for multiple comparisons to explore post hoc differences between pairs of groups.

Multiple logistic regression analysis was performed, entering explanatory variables that showed significant results in univariate tests (P < 0.05) or were considered a priori to be of foremost importance for the study outcome, with ANA positivity as the outcome variable. Cases with missing variables were excluded from the analysis. Before the application of logistic regression procedures, some continuous variables were dichotomized to binary variables. For age at disease presentation, the cut points chosen were ≤6 years and >6 years. Cut points for the cumulative number of joints affected at 6, 12, and 24 months were obtained through receiver operating characteristic (ROC) curve analysis. The step-down strategy of analysis was chosen; this consists of examining the effect of removing variables from the saturated model. Possible explanatory variables assessed were sex, age at disease presentation, symmetric arthritis at disease presentation and at 6 months, iridocyclitis, cumulative number of affected joints at 6, 12, and 24 months, radiographic changes, and type of joints involved in the first 24 months after disease presentation (shoulder, elbow, wrist, metacarpophalangeal, proximal interphalangeal, hip, knee, ankle, foot, and finger joints). The effect was expressed in terms of odds ratios, and 95% confidence intervals were calculated; statistical significance was tested by likelihood ratio test. The area under the ROC of the best-fitting model was used as an indicator of the predictive ability of the model.

The association between categorical variables was further explored by multiple correspondence analysis (18). Multiple correspondence analysis is an exploratory analysis and descriptive technique that enables analysis of contingency tables with a large number of variables, considering measures of correspondence between rows and columns. It can be considered a variant of factor analysis, but it differs because it evaluates the relationship between categorical data organized in contingency tables, rather than continuous data. Essentially, multiple correspondence analysis converts a non-negative data matrix into a kind of graphic representation that allows studying the relationship between the categories in a multiway contingency table. The importance of the categories of variables for constructing each axis is measured by their absolute contribution and aids in the interpretation of the axis. Categories with high absolute contribution show greater importance in the factor's formation. The relative contribution provides information as to how much of the category's variability is explained by the axis. Graphic analysis of association of variables is performed by considering their geometric proximity and the separation of categories by quadrants, because the closer the variables, the more interrelated they are, and those separated by quadrants display groups with opposite profiles. The variables that proved statistically significant in multivariate analysis were used in the multiple correspondence analysis. Only patients for whom complete information was available were included.

All statistical tests were 2-sided; P values less than 0.05 were considered significant. The statistical packages used were Statistica (version 8.0; StatSoft) for univariate analyses and Stata release 11 (StataCorp) for multivariate and cluster analysis. For multiple correspondence analysis, the software XLSTAT, 6.1.9 (Addinsoft) was used.

RESULTS

A total of 1,219 patients fulfilled the ILAR criteria for JIA. The ILAR category was systemic arthritis in 149 patients (12.2%), oligoarthritis in 649 patients (53.2%), RF-positive polyarthritis in 26 patients (2.1%), RF-negative polyarthritis in 223 patients (18.3%), enthesitis-related arthritis in 73 patients (6.0%), psoriatic arthritis in 37 patients (3.0%), and undifferentiated arthritis in 62 patients (5.1%). Of the 649 patients with oligoarthritis, 433 had the persistent subtype and 195 had the extended subtype; 21 patients could not be subclassified as having the persistent or extended subtype due to a disease duration of <6 months. For the above-mentioned reasons, patients with systemic arthritis, RF-positive polyarthritis, and enthesitis-related arthritis were excluded from the study. The remaining 971 patients in the ILAR categories of oligoarthritis, RF-negative polyarthritis, psoriatic arthritis, and undifferentiated arthritis were combined and classified according to their ANA status, as follows: 711 (73.2%) were ANA positive, 149 (15.3%) were ANA negative, and 111 (11.4%) had an indeterminate ANA status. Patients with an indeterminate ANA status were excluded from the analysis. The number of ANA determinations per patient in the 860 patients who had the ANA status specified ranged from 2 to 20 (mean 5.4); the total number of determinations was 4,610. More than 90% of patients were of Italian ancestry.

Table 1 shows the comparison of demographic and clinical features, cumulative joint involvement over time, and frequency of involvement of specific joints in the first 24 months after disease presentation between ANA-positive and ANA-negative patients. The 2 patient groups were significantly different for all features, except for the frequency of involvement of ankle and small foot joints. Compared with the ANA-negative group, the ANA-positive patients had a higher proportion of females, a younger age at disease presentation, a greater prevalence of iridocyclitis and of asymmetric arthritis at disease presentation and in the first 6 months, a lower frequency of HLA–B27, a lower number of affected joints over time, a lower frequency of involvement of the shoulder, elbow, wrist, small hand joints, and hip, a greater frequency of involvement of the knee, and a lower frequency of radiographic joint changes. The duration between disease presentation and last followup visit was comparable in the 2 patient populations.

Table 1. Main demographic and clinical features, cumulative joint involvement over time, and frequency of involvement of specific joints in the first 24 months after disease presentation in the 860 patients with juvenile idiopathic arthritis, according to ANA status*
 ANA positive (n = 711)ANA negative (n = 149)P
  • *

    Except where indicated otherwise, values are the number/number tested (%). ANA = antinuclear antibody; IQR = interquartile range.

  • By chi-square test unless otherwise specified.

  • By Mann-Whitney U test.

  • §

    By Fisher's exact test.

Female sex579/711 (81.43)93/149 (62.42)<0.0001
Age at disease presentation, median (IQR) years2.7 (1.7–5.2)6.3 (3.6–9.7)<0.0001
Age ≤6 years at disease presentation564/711 (79.32)72/149 (48.32)<0.0001
Iridocyclitis ever186/709 (26.23)4/149 (2.68)<0.0001
Presence of HLA–B2718/477 (3.8)11/121 (9.1)0.015§
Asymmetric arthritis at disease presentation605/696 (86.93)94/136 (69.12)<0.0001
Asymmetric arthritis at 6 months538/682 (78.65)80/136 (58.82)<0.0001
Cumulative no. of joints affected at 6 months, median (IQR)2.0 (1.0–4.0)3.0 (1.0–9.0)0.0026
Cumulative no. of joints affected at 12 months, median (IQR)3.0 (2.0–5.0)4.0 (2.0–13.5)0.0005
Cumulative no. of joints affected at 24 months, median (IQR)3.0 (2.0–7.0)5.0 (2.0–16.0)0.0004
Joints affected in the first 24 months   
 Shoulder25/706 (3.54)20/148 (13.51)<0.0001
 Elbow124/706 (17.56)39/148 (26.35)0.013
 Wrist175/707 (24.75)70/148 (47.30)<0.0001
 Metacarpophalangeal137/706 (19.41)46/148 (31.08)0.002
 Proximal interphalangeal210/704 (29.83)61/148 (41.22)0.007
 Hip45/705 (6.38)44/148 (29.73)<0.0001
 Knee634/707 (89.67)120/148 (81.08)0.003
 Ankle420/707 (59.41)85/148 (57.43)0.65
 Metatarsophalangeal65/704 (9.23)15/148 (10.14)0.73
 Foot interphalangeal81/706 (11.47)18/148 (12.16)0.81
Presence of radiographic joint changes98/496 (19.8)53/133 (39.8)<0.0001§
Disease duration at last followup visit, median (IQR) years5.0 (2.6–8.4)5.1 (1.8–9.4)0.45

The comparison of the main demographic and clinical features of ANA-positive and ANA-negative patients by ILAR category is presented in Table 2. Overall, the differences noted when the 2 patient groups were considered as a whole were paralleled by their comparison within each category. The sole exception was the similar frequency of asymmetric arthritis in ANA-positive and ANA-negative patients with persistent oligoarthritis and psoriatic arthritis. ANA-positive patients with psoriatic arthritis and undifferentiated arthritis tended to be less often female and to be older at disease presentation than ANA-positive patients with oligoarthritis (either persistent or extended) and RF-negative polyarthritis. In the categories extended oligoarthritis, RF-negative polyarthritis, and undifferentiated arthritis, the cumulative number of joints affected in the first 24 months after disease presentation was lower in ANA-positive patients than in ANA-negative patients. The cumulative number of joints involved over time was comparable between ANA-positive and ANA-negative patients in the categories of persistent oligoarthritis and psoriatic arthritis.

Table 2. Main demographic and clinical features by ILAR category and ANA status*
 Persistent oligoarthritisExtended oligoarthritisRF-negative polyarthritisPsoriatic arthritisUndifferentiated arthritis
ANA+ (n = 326)ANA– (n = 56)ANA+ (n = 160)ANA– (n = 12)ANA+ (n = 148)ANA– (n = 49)ANA+ (n = 25)ANA– (n = 8)ANA+ (n = 36)ANA– (n = 19)
  • *

    Except where indicated otherwise, values are the number/number tested (%). Data on 21 patients with nonspecified oligoarticular arthritis were excluded from this table. Comparisons of quantitative data were made by Mann-Whitney U test; comparisons of frequencies were made by chi-square test (or by Fisher's exact test if expected frequencies were <5). ILAR = International League of Associations for Rheumatology; ANA = antinuclear antibody; RF = rheumatoid factor; IQR = interquartile range.

  • P < 0.001 versus ANA-negative patients.

  • P < 0.05 versus ANA-negative patients.

  • §

    P < 0.0001 versus ANA-negative patients.

  • P < 0.01 versus ANA-negative patients.

Female sex268/326 (82.2)34/56 (60.7)138/160 (86.2)8/12 (66.67)119/148 (80.4)33/49 (67.3)17/25 (68.0)2/8 (25.0)24/36 (66.7)11/19 (57.9)
Age at disease presentation, median (IQR) years2.8 (1.9–5.0)§5.4 (2.9–9.0)2.5 (1.6–4.6)6.8 (4.1–9.0)2.4 (1.6–6.1)§6.4 (4.4–9.1)3.5 (2.0–7.1)8.3 (5.2–12.7)3.4 (1.7–8.1)7.4 (2.8–10.2)
Asymmetric arthritis at 6 months274/319 (85.9)45/52 (86.5)120/157 (76.4)5/11 (45.4)94/143 (65.7)§13/42 (30.9)15/21 (71.4)7/8 (87.5)25/32 (78.1)6/18 (33.3)
Iridocyclitis ever78/326 (23.9)1/56 (1.8)43/160 (26.9)0/12 (0.0)43/148 (29.0)2/49 (4.1)7/25 (28.0)0/8 (0.0)9/35 (25.7)1/19 (5.3)
Cumulative no. of joints affected at 24 months, median (IQR)2 (1–3)2 (1–3)5 (3–7)10 (6–20.5)10 (7–14)§17 (9–26)5 (2–10)4 (2.0–8.5)3 (2–5)9 (2–23)
Joints affected in the first 24 months          
 Wrist19/326 (5.8)6/56 (10.7)52/160 (32.5)8/12 (66.7)85/148 (57.4)42/49 (85.7)11/25 (44.0)5/8 (62.5)7/3 (19.4)9/19 (47.4)
 Metacarpophalangeal8/326 (2.4)1/56 (1.8)45/160 (28.1)7/12 (58.3)68/148 (45.9)27/49 (55.1)8/25 (68.0)3/8 (37.5)8/36 (22.2)8/19 (42.1)
 Proximal interphalangeal24/326 (7.4)5/56 (8.9)64/158 (40.5)7/12 (58.3)97/148 (65.5)38/49 (77.5)13/25 (52.0)2/8 (25.0)12/36 (33.3)9/19 (47.4)
 Hip8/326 (2.4)6/56 (10.7)14/159 (8.8)5/12 (41.7)20/148 (13.5)§23/49 (46.9)2/25 (8.0)1/8 (12.5)1/36 (2.8)§9/19 (47.4)
 Knee296/326 (90.8)46/56 (82.1)146/160 (91.2)11/12 (91.7)137/148 (92.6)38/49 (77.5)19/25 (76.0)4/8 (50.0)26/36 (72.2)17/19 (89.5)
 Ankle130/326 (39.9)19/56 (33.9)114/160 (71.2)10/12 (83.3)131/148 (88.5)42/49 (85.7)13/25 (52.0)4/8 (50.0)25/36 (69.4)10/19 (52.6)

Compared with their ANA-negative counterparts, ANA-positive patients in all categories had a lower frequency of involvement of the wrist, small hand joints, and hip, and a greater frequency of involvement of the knee. Exceptions were the comparable frequency of involvement of the small hand joints in ANA-positive and ANA-negative patients with persistent oligoarthritis, the greater frequency of involvement of the same joints in ANA-positive than in ANA-negative patients with psoriatic arthritis, the comparable frequency of knee involvement in ANA-positive and ANA-negative patients with extended oligoarthritis, and the greater frequency of knee involvement in ANA-negative than in ANA-positive patients with undifferentiated arthritis. Data regarding involvement of the distal interphalangeal joints were not available. The disease duration was comparable between ANA-positive and ANA-negative patients within each category (results not shown). The values of the same demographic and clinical features for the ANA-positive and ANA-negative patients combined for each JIA category are provided in Table 3. The comparison of these data with those in Table 2 highlights the relative homogeneity of ANA-based grouping compared with grouping according to the ILAR classification.

Table 3. Main demographic and clinical features by ILAR category*
 Persistent oligoarthritis (n = 382)Extended oligoarthritis (n = 172)RF-negative polyarthritis (n = 197)Psoriatic arthritis (n = 33)Undifferentiated arthritis (n = 55)PComparisons significant on post hoc tests
  • *

    Except where indicated otherwise, values are the number/number tested (%). Data on 21 patients with nonspecified oligoarticular arthritis were excluded from this table. Comparisons of quantitative data were made by Mann-Whitney U test; comparisons of frequencies were made by chi-square test (or by Fisher's exact test if expected frequencies were <5). ILAR = International League of Associations for Rheumatology; IQR = interquartile range.

  • For overall comparisons.

  • Pairs of comparisons that were statistically significant on post hoc tests (Bonferroni adjustment or Dunn's test). PsA = psoriatic arthritis; OP = persistent oligoarthritis; OE = extended oligoarthritis; UA = undifferentiated arthritis; PO = rheumatoid factor (RF)–negative polyarthritis.

Female sex302/382 (79.06)146/172 (84.88)152/197 (77.16)19/33 (57.58)35/55 (63.64)0.0006PsA vs. OP and OE, UA vs. OE
Age at disease presentation, median (IQR) years4.30 (1.91–5.61)2.72 (1.66–4.85)4.75 (1.77–6.98)4.52 (2.33–8.33)4.65 (2.29–9.55)0.0008OE vs. PsA and UA
Asymmetric arthritis at 6 months52/371 (14.02)43/168 (25.60)78/185 (42.16)7/29 (24.14)19/50 (38.00)<0.0001PO vs. OP and OE, UA vs. OE, OP vs. OE
Iridocyclitis ever79/382 (20.68)43/172 (25.00)45/197 (22.84)7/33 (21.21)10/54 (18.52)0.78
Cumulative no. of joints affected at 24 months, median (IQR)2.0 (1.0–3.0)5.0 (4.00–7.0)10.0 (7.0–16.0)5.00 (2.0–10.0)4.00 (2.0–11.0)<0.0001PO vs. OP, OE, PsA, and UA; OP vs. PsA, UA, and OE
Joints affected in the first 24 months       
 Wrist25/382 (6.54)60/172 (34.88)127/197 (64.47)16/33 (48.48)16/55 (29.09)<0.0001PO vs. OE, OP, and UA; OP vs. UA
 Metacarpophalangeal9/382 (2.36)52/172 (30.23)95/197 (48.22)11/33 (33.33)16/55 (29.09)<0.0001PO vs. OP and OE, OP vs. OE
 Proximal interphalangeal29/382 (7.59)71/170 (41.76)135/197 (68.53)15/33 (45.45)21/55 (38.18)<0.0001PO vs. UA, OP, and OE; UA vs. OP
 Hip14/382 (3.66)19/171 (11.11)43/197 (21.38)3/33 (9.09)10/55 (18.18)<0.0001PO vs. OP, OP vs. OE
 Knee342/382 (89.53)157/172 (91.28)175/197 (88.83)23/33 (69.70)43/55 (78.18)0.001PsA vs. OP and OE
 Ankle149/382 (39.01)124/172 (72.09)173/197 (87.82)17/33 (51.52)35/55 (63.64)<0.0001PO vs. PsA, UA, OP, and OE; UA vs. OP; OP vs. OE

For the multivariate analysis, complete data were available on 785 patients. The best-fitting model obtained through logistic regression procedures, in which the presence of ANAs was the dependent variable, is presented in Table 4. Independent correlations with the presence of ANAs were identified for presence or history of iridocyclitis, female sex, age ≤6 years at disease presentation, cumulative affected joints at 12 months ≤12, presence of asymmetric arthritis at 6 months, and absence of hip joint involvement. To investigate whether use of a different cutoff for ANA positivity would lead to different results, we repeated the regression analysis by adding 74 patients who were ANA positive at low titer (i.e., who had at least 2 positive ANA determinations at a titer ≥1:40) and for whom complete information was available. Compared with the original analysis, the variables presence of wrist joint involvement and presence of metacarpophalangeal joint involvement entered the best-fitting model, whereas the variable presence of asymmetric arthritis at 6 months did not (results not shown).

Table 4. Best-fitting model obtained through logistic regression procedures*
Explanatory variableOR (95% CI)P
  • *

    The presence of antinuclear antibodies was the dependent variable. Complete data were available on 785 patients. The area under the receiver operating characteristic curve of the model was 0.82. OR = odds ratio; 95% CI = 95% confidence interval.

  • By likelihood ratio test.

Iridocyclitis ever22.03 (5.09–95.29)<0.0001
Female sex2.67 (1.69–4.21)<0.0001
Age ≤6 years at disease presentation2.46 (1.61–3.77)<0.0001
Cumulative affected joints at 12 months ≤123.16 (1.72–5.79)0.0003
Asymmetric arthritis at 6 months1.69 (1.06–2.70)0.03
Presence of hip joint involvement0.33 (0.19–0.57)0.0001

The association between the presence of ANA and the variables that proved statistically significant in multivariate analysis was further explored by means of multiple correspondence analysis. For the purposes of this analysis, all variables were categorized as was done in logistic regression procedures. Figure 1 depicts the 2-dimensional representation of the variables examined. This figure highlights the existence of 2 groups. One group, located in the upper and lower left quadrants, whose proximity between categories includes ANA positivity, is characterized by age at disease presentation ≤6 years, greater prevalence of females, asymmetric arthritis, lower number of affected joints over time, lack of hip involvement, and presence of iridocyclitis. A second group, located in the upper and lower right quadrants, whose proximity between categories includes ANA negativity, is characterized by greater prevalence of males, age at disease presentation >6 years, symmetric arthritis, greater number of affected joints over time, presence of hip involvement, and lack of iridocyclitis.

Figure 1.

Graphic representation of variable categories in the first 2 dimensions of multiple correspondence analysis. ANA = antinuclear antibody.

DISCUSSION

The results of this analysis show that ANA-positive patients classified into different disease categories (persistent oligoarthritis, extended oligoarthritis, RF-negative polyarthritis, psoriatic arthritis, and undifferentiated arthritis) by the current JIA criteria share similar characteristics (e.g., strong predominance of females, early onset of disease, asymmetric arthritis, and high risk of chronic iridocyclitis). This confirms and expands our previous findings (16) and substantiates our original hypothesis that ANA-positive patients with JIA constitute a homogeneous subgroup, irrespective of the course of joint disease and the presence of psoriatic features (15).

As compared to our former study (16), the present investigation included a much larger patient sample and extended the analysis to the psoriatic and undifferentiated arthritis categories. We found that ANA-positive patients had a lower number of affected joints over time and a lower frequency of involvement of some joints and of radiographic joint lesions than did ANA-negative patients. These differences were seen when ANA-positive and ANA-negative patients were compared either as whole groups or within the single categories. The diversities between ANA-positive and ANA-negative patients were confirmed by multivariate regression analysis and were further highlighted by the 2-dimensional scatterplot of multiple correspondence analysis.

Our findings further underscore the need to reconsider some of the JIA categories. Should the ILAR classification be used, the study patients would be placed in their respective category based on the number of joints affected over time or the presence of psoriasis or psoriatic features. As one of us previously discussed in detail elsewhere (15), the total number of joints involved may not be an appropriate classification tool, as it may simply reflect a more rapid spread of arthritis within the same disease. Previous studies have demonstrated that the reliability of clinical examination of joints in children with JIA is poor (19). A recent analysis has shown a high prevalence of subclinical synovitis as detected by ultrasound in children with JIA. Some patients who were labeled as having oligoarthritis or who were found to have no synovitis on clinical evaluation were subsequently determined with the use of ultrasound to have polyarthritis (20).

The heterogeneity of RF-negative polyarthritis has been highlighted recently by gene expression studies. Griffin et al (21) found 3 distinct gene expression signatures with variable expression among patients with polyarticular JIA, with signature I identifying patients with a disease similar to adult rheumatoid arthritis, signature II not being associated with any disease subset, signature III identifying patients with less inflammatory disease, and patients with ANA-positive early-onset disease expressing none of these signatures.

There is evidence, as well, that juvenile psoriatic arthritis is not a unique entity. Two distinct subtypes of this condition have been recognized (15). One is very similar to early-onset ANA-positive oligoarthritis, although often with a more rapid spread of arthritis; the other shares the features of enthesitis-related arthritis and therefore belongs to the group of spondylarthritides. In accordance with this hypothesis, Stoll et al (22) recently identified 2 distinct populations of patients with juvenile psoriatic arthritis diagnosed according to the Vancouver criteria (23): one group with younger age at onset, a greater prevalence of females, and more common expression of ANAs, and another group with older age at onset, an even number of boys and girls, and an increased incidence of axial disease and enthesitis. In the ILAR classification (as opposed to the Vancouver criteria), patients with enthesitis are excluded from the category of psoriatic arthritis. Indeed, most of our patients with psoriatic arthritis had the features of ANA-positive oligoarthritis.

The choice of a family history of psoriasis as an exclusion criterion (leading to placement of a patient in the undifferentiated arthritis category) in the ILAR classification has been a matter of controversy (14, 24). As many as 82% of our patients who fell into the undifferentiated arthritis category were classified as such due to the presence of a family history of psoriasis in a first-degree relative. We previously found that a positive family history of psoriasis does not affect the clinical picture and course in JIA patients with oligoarthritis, which contradicts the use of such a history as an exclusion criterion (25). In accordance with this criticism, we found that ANA-positive patients in the undifferentiated arthritis category were comparable to ANA-positive patients in the other ILAR subsets.

A number of potential limitations to the study must be acknowledged, the most important of which is its retrospective nature. Retrospective data collection is subject to missing and possibly erroneous data. In the absence of agreed-upon criteria for positivity of ANA, we relied on tests performed on rodent or HEp-2 substrates (13) and chose a cutoff of 1:160 to define positivity. However, in ∼11% of our patients, ANA status could not be defined by these stringent criteria. Furthermore, although in the majority of ANA determinations in patients classified as ANA positive the titer was ≥1:160, some repeat determinations in these patients were positive at a lower titer or even negative. We recognize that our choice of the 1:160 cutoff is arbitrary and that other cutoffs may prove equally suitable in other centers or laboratories. A multinational collaborative effort is necessary to establish the optimal threshold for ANA positivity in children with JIA. The frequency of ANA positivity in our series (73.2%) may appear high. However, this frequency was assessed only in select categories of JIA. Considering all categories, the proportion of ANA-positive patients in our series drops to 59.7%, which is in accordance with that recently reported in a large sample of JIA patients followed up in a Western tertiary care pediatric rheumatology center (55.5%) (26). However, because our study population was almost exclusively of Italian ancestry, it is not certain if this ANA-positive phenotype is similarly seen in all genetic backgrounds.

The absence of RF was confirmed in most RF-negative patients with at least 2 determinations. However, we cannot ensure that the RF test was repeated many times in patients who were initially RF negative. Therefore, we cannot exclude the possibility that some patients who were initially RF negative could have become RF positive over time. Finally, we should acknowledge that our definition of joint symmetry is arbitrary. An agreed-upon definition of symmetry requires consensus among international expert pediatric rheumatologists. The main strength of our study lies in the large number of patients enrolled and in the thorough clinical and statistical analysis.

In conclusion, our findings confirm our previous hypothesis that patients with a distinct set of features, including ANA positivity, young age at disease onset, female sex, asymmetric arthritis, and high risk of developing chronic iridocyclitis, represent a homogeneous group, irrespective of the course of joint disease or the presence of psoriatic features. Further support for our hypothesis comes from the recent observation that gene expression differences, in part representing a B cell signature, are able to distinguish early- and late-onset JIA, independently of the number of joints involved (27).

We therefore suggest the need to revise some of the present JIA classification categories in order to better identify homogeneous patient populations for future genetic and immunopathogenetic investigations, outcome studies, and clinical trials. This effort should be paralleled by an investigation of whether the ANA-defined subgroups are distinct with regard to objective biomarkers, preferably ones that are relevant to disease etiology.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Ravelli had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ravelli, Ruperto, Martini.

Acquisition of data. Varnier, Oliveira, Castell, Arguedas, Magnani, Magni-Manzoni, Lattanzi, Dalprà, Battagliese, Verazza, Allegra.

Analysis and interpretation of data. Ravelli, Pistorio, Ruperto, Galasso, Martini.

Ancillary