ARTHRITIS & RHEUMATISM Vol. 54, No. 11, November 2006, pp 3564–3572 DOI 10.1002/art.22173 © 2006, American College of Rheumatology Patients With Juvenile Psoriatic Arthritis Comprise Two Distinct Populations Matthew L. Stoll,1 David Zurakowski,1 Lise E. Nigrovic,1 David P. Nichols,2 Robert P. Sundel,1 and Peter A. Nigrovic3 than twice as long to achieve clinical remission (23 months versus 9.2 months; P ⴝ 0.044). Cluster analysis identified largely overlapping subgroups but suggested that the presence of dactylitis, rather than age, has the greatest capacity to predict essential features of the clinical phenotype. Conclusion. Juvenile PsA comprises 2 distinct populations of patients. Although the pathophysiologic correlate of this finding remains undefined, future studies should avoid the assumption that PsA in childhood constitutes a single etiologic entity. Objective. Psoriatic arthritis (PsA) in children is clinically heterogeneous. We examined a large population of children with juvenile PsA for evidence of phenotypic clustering that could suggest the presence of distinct clinical entities. Methods. We reviewed the medical records of 139 patients meeting the Vancouver criteria for juvenile PsA. To identify segregation into phenotypic groups, we compared younger patients with their older counterparts and subjected the whole population to 2-step cluster analysis. Results. Among patients with juvenile PsA, the age at onset is biphasic, with peaks occurring at approximately 2 years of age and again in late childhood. Compared with children ages 5 years and older, younger patients are more likely to be female, exhibit dactylitis and small joint involvement, and express antinuclear antibodies. Progression to polyarticular disease (>5 joints) is more common in younger children, although joint involvement remains oligoarticular in the majority of children. In contrast, older patents tend to manifest enthesitis, axial joint disease, and persistent oligoarthritis. Uveitis is equally represented in both age groups. Despite a higher utilization of methotrexate therapy, younger patients required, on average, more The classification of pediatric rheumatic disease is hampered by a limited understanding of the pathogenesis. Hence, patients have traditionally been grouped according to clinical disease pattern; such grouping has important implications for research and clinical care. One such diagnostic grouping is juvenile psoriatic arthritis (PsA). Juvenile PsA was initially described in children with arthritis who at some point in their disease course exhibited frank psoriasis (1–5); however, it is now well accepted that juvenile PsA may be diagnosed in patients without the classic skin rash but meeting other criteria, including typical skin/nail changes and a positive family history of psoriasis (6–9). Using this standard, juvenile PsA is diagnosed in ⬃5% of children with arthritis seen in pediatric rheumatology clinics, a prevalence slightly higher than that of rheumatoid factor–positive juvenile polyarticular arthritis (10). Both the Vancouver criteria for juvenile PsA (6) and the International League of Associations for Rheumatology (ILAR) nomenclature for juvenile idiopathic arthritis (JIA) (8) consider PsA in children to represent a single clinical entity. In several published series, however, clinical and genetic differences have been observed between younger patients and older patients, although small patient numbers have prohibited formal analysis for subpopulations (6,9,11). Recognition of this heterogeneity resulted in a call for a more detailed clinical Supported in part by the Samara Jan Turkel Center for Pediatric Autoimmune Disease. Dr. Lise E. Nigrovic’s work was supported by a National Research Service Award grant (T32 HD40128). 1 Matthew L. Stoll, MD, PhD, David Zurakowski, PhD, Lise E. Nigrovic, MD, MPH, Robert P. Sundel, MD: Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts; 2David P. Nichols, AM: University of Chicago and SPSS Inc., Chicago, Illinois; 3 Peter A. Nigrovic, MD: Children’s Hospital Boston, Harvard Medical School, and Brigham and Women’s Hospital, Boston, Massachusetts. Mr. Nichols owns stock and/or stock options in SPSS. Address correspondence and reprint requests to Peter A. Nigrovic, MD, Rheumatology, Children’s Hospital Boston, 300 Longwood Avenue, Fegan 6, Boston, MA 02115. E-mail: [email protected] partners.org. Submitted for publication March 13, 2006; accepted in revised form July 24, 2006. 3564 TWO POPULATIONS OF PATIENTS WITH JUVENILE PsA Table 1. Vancouver criteria for juvenile psoriatic arthritis* Major criteria Arthritis Psoriasis Minor criteria Nail pitting Dactylitis Family history of psoriasis in a first- or second-degree relative Psoriasis-like lesion * Definite juvenile psoriatic arthritis (PsA) ⫽ 2 major criteria, or arthritis plus 3–4 minor criteria; probable juvenile PsA ⫽ arthritis plus 2 minor criteria (for review, see ref. 6). subgrouping within juvenile PsA, but more than a decade later this has not been accomplished (11). The identification of subgroups would be of considerable interest, because it would enable more accurate phenotypic classification of patient groups for the purposes of scientific investigation. Furthermore, distinct patient groups might be expected to respond differentially to treatment, potentially permitting more accurate therapeutic decision-making. Accordingly, we studied our patients with juvenile PsA to determine whether they constitute a single population or multiple populations. Here, we report that juvenile PsA constitutes 2 distinct patient subgroups that are distinguished by age at disease onset and the presence of dactylitis and are differentiated by sex ratio, joints affected, laboratory values, and clinical course. These findings suggest that current diagnostic criteria capture 2 disease entities and indicate that PsA in children is more complex than previously appreciated. PATIENTS AND METHODS Patients. We reviewed the charts of every patient seen at the rheumatology clinic of Children’s Hospital Boston between January 1997 and February 2005, for whom the International Classification of Diseases, Ninth Revision codes were 696.1 (psoriasis), 696.0 (PsA), or 756.11 (spondylarthritis). Patients were included in the study if they fulfilled the Vancouver criteria for probable or definite juvenile PsA (6) (Table 1). These criteria were selected instead of the ILAR criteria for psoriatic JIA (8), because many patients met criteria on the basis of rashes that were judged as being psoriasiform but were not formally diagnosed as psoriasis; such a presentation is especially common in younger children (12). We reviewed the medical records for pertinent historic elements, physical examination findings, and laboratory values. When available, studies obtained elsewhere that were contained in the medical record were included in the review. The following definitions were used to categorize findings from chart review. A patient was considered to have psoriasis if that diagnosis had been given definitively by a 3565 dermatologist or other physician; rashes noted by the examining rheumatologist and thought likely (but not definitively) to represent psoriasis were considered psoriasis-like. Polyarticular arthritis was defined by the involvement of ⱖ5 joints cumulatively at any point over the course of observation. Oligoarticular arthritis was defined by involvement of ⬍5 joints. Small peripheral joints were considered to be the metacarpophalangeal and interphalangeal joints of the hands and the corresponding joints of the feet. Large peripheral joints included the wrists, elbows, knees, and ankles. Axial joints included the temporomandibular joints, shoulders, cervical or lumbar spine, sacroiliac joints, and hips. Dactylitis was defined as digital swelling extending beyond the margins of the joint. All patients with dactylitis were considered also to have small-joint arthritis in the corresponding digit(s), although specific attribution to proximal interphalangeal or distal interphalangeal joint involvement could not always be made. Enthesitis was defined as any tenderness at tendinous, ligamentous, capsular, or fascial insertions into bone, as determined by the examining attending pediatric rheumatologist. Remission was defined as the absence of clinically evident synovitis or enthesitis, on or off medications; laboratory parameters were not employed. Remission off medications was defined as remission in any patient not actively receiving therapy, excluding nonsteroidal antiinflammatory drugs taken as needed but less often than daily. Antinuclear antibody (ANA) values were considered positive if they were above the upper limits of normal for the laboratory in which the test was performed. A Steinbrocker class was assigned by the rheumatologist reviewing the chart, on the basis of functional restrictions reported at the final visit (13). Institutional review board approval was obtained for this study. Statistical analysis. We compared categorical data and proportions using the chi-square test or Fisher’s exact test, as indicated. Means were compared with Student’s t-test, and medians were compared with the Mann-Whitney U test. Analysis of normality was performed with the KolmogorovSmirnov test with Lilliefors correction, with a significant P value indicating evidence of non-normality. Because this was an exploratory study, we elected to display differences significant at a 2-tailed P value of less than 0.05, uncorrected for multiple comparisons (14,15). The more conservative Bonferroni correction was also applied, and the resulting significance thresholds are shown in the footnotes of the respective tables. To analyze our data for clinical subsets, we applied 2-step cluster analysis using log-likelihood distance measures. For this analysis, we selected the following 8 clinical and laboratory variables: sex, presence of psoriasis, dactylitis, enthesitis, axial disease, oligoarticular onset, and ANA positivity (all expressed as categorical variables), with age at onset as a continuous variable. Only patients for whom complete information was available were included. The goal was to identify clusters that minimize differences within groups and maximize differences between groups. The first step involved a preclustering routine that arrayed patients into a cluster feature tree. The initial patient defined a subcluster, and each additional patient was added either to the existing subcluster or to a new one, depending on the degree of similarity to patients in existing subclusters. In the second step, subclusters within the cluster feature tree were grouped using an agglomerative hierarchical clustering algorithm designed to identify the optimal number of clusters (from 1 to 15) based on functions of the 3566 STOLL ET AL Table 2. Clinical characteristics of patients according to subclass of juvenile PsA* Characteristic Female sex, % Duration of followup, median (IQR) months Psoriasis, % Age, median (IQR) years Joints affected, % Any axial Temporomandibular Shoulder Spine Sacroiliac Hip Any peripheral large joint Elbow Wrist Knee Ankle Any peripheral small joint Metacarpophalangeal Proximal interphalangeal Distal interphalangeal Metatarsophalangeal Dactylitis, % Enthesitis, % ESR, mean ⫾ SD mm/hour C-reactive protein, mean ⫾ SD mg/dl Platelet count, ⫻1,000 cells/l, mean ⫾ SD White blood cell count, ⫻1,000 cells/l, mean ⫾ SD ANA status, no. positive/no. tested (%) HLA–B27 status, no. positive/no. tested (%) Treatment, % Sulfasalazine Methotrexate Tumor necrosis factor inhibitors Any disease-modifying antirheumatic drug Disease course, % Oligoarticular at first visit Oligoarticular at 6 months† Oligoarticular course† Uveitis ever, no. positive/no. tested (%) Remission off therapy at last examination, %‡ Remission at last examination, %‡ Time to remission, median (IQR) months‡ Steinbrocker class I at last visit, %‡ Total (n ⫽ 139) Probable PsA (n ⫽ 73) Definite PsA (n ⫽ 66) 59 23 (8.5–49) 25 7.3 (2.5–11) 49 23 (9.1–42) 0 7.8 (2.6–10) 70 23 (6.2–55) 53 7.2 (2.3–12) 20 6.5 2.9 2.2 0.7 11 80 13 25 60 51 57 18 24 9.4 13 37 45 25 ⫾ 20 0.6 ⫾ 1 361 ⫾ 102 8.7 ⫾ 3.7 54/117 (46) 6/13 (46) 25 6.8 5.5 0 0 14 77 14 27 58 47 56 15 25 9.6 11 32 49 26 ⫾ 24 0.6 ⫾ 0.8 365 ⫾ 103 8.8 ⫾ 4.1 29/62 (47) 3/7 (43) 15 6.1 0 4.5 1.5 7.6 83 12 23 64 56 58 21 24 9.1 15 42 39 25 ⫾ 15 0.6 ⫾ 1 356 ⫾ 102 8.5 ⫾ 3.3 25/55 (45) 3/6 (50) 52 46 12 74 89 84 77 6/76 (7.9) 41 57 15 (4.1–42) 81 47 45 8.2 68 92 87 82 3/46 (6.5) 48 58 16 (5.3–36) 79 58 47 17 80 86 80 71 3/30 (10) 32 56 15 (3.0–48) 84 P 0.015 NS ⬍0.001 NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS NS * Except where indicated otherwise, values are the mean ⫾ SD. Probable versus definite juvenile psoriatic arthritis (PsA) was defined according to the Vancouver criteria (see Table 1). P values less than 0.05 were considered significant. Bonferroni’s adjustment changed the significance threshold to P ⬍ 0.0013. IQR ⫽ interquartile range; NS ⫽ not significant; ESR ⫽ erythrocyte sedimentation rate; ANA ⫽ antinuclear antibody. † Among those with at least 6 months of followup (for probable PsA, n ⫽ 60; for definite PsA, n ⫽ 51). ‡ Reported for all patients with at least 2 visits (for probable PsA, n ⫽ 68; for definite PsA, n ⫽ 61). Schwarz-Bayesian information criterion. Because the order of patient entry into the cluster algorithm can affect the clusters formed, we ran the analysis 100 times with randomly scrambled patient order to assess the robustness of cluster findings (16). Statistical analysis was performed using SPSS software (version 13.0; SPSS, Chicago, IL). RESULTS A total of 139 patients fulfilled the Vancouver criteria for juvenile PsA (Table 2). The median age at disease onset was 7.3 years, and 59% of the patients were female. The duration of followup ranged from 1 visit to 8 years (median 23 months). Patients had between 1 and 33 visits (median 6). Among the 139 patients, psoriasis was diagnosed in 35 at some point over the course of observation. Among the 104 children without psoriasis, 73 (53% of the total group) met exactly 2 supplemental criteria, thus fulfilling the diagnosis of probable PsA, while 31 chil- TWO POPULATIONS OF PATIENTS WITH JUVENILE PsA Figure 1. Age at onset of juvenile psoriatic arthritis. The data for age at onset are inconsistent with a single normal distribution (P ⬍ 0.001 by Kolmogorov-Smirnov test with Lilliefors correction), conforming best to 2 distributions (age at onset younger than age 5 years and 5 years of age and older [median ages 1.8 years and 10 years, respectively]). dren met 3 or more supplemental criteria. These 31 patients, plus the 35 patients with psoriasis, constituted the group with definite juvenile PsA. As shown in Table 2, aside from the group of patients with definite juvenile PsA having a higher percentage of psoriasis (53% versus 0% in the group with probable juvenile PsA; P ⬍ 0.001) and females (70% versus 49% in the group with probable juvenile PsA; P ⫽ 0.015), there were no differences in the demographic, clinical, or treatment parameters between the group with definite juvenile PsA and the group with probable juvenile PsA. Thus, the patients were grouped for all subsequent data analysis. The distribution of joint involvement is shown in Table 2. In total, 28 children (20%) had some form of axial disease. Of the larger joints, the knee was involved most commonly, followed by the ankle, wrist, and elbow. More than one-half of the patients exhibited synovitis of the small joints of the hands or feet, including dactylitis in more than one-third of patients. Enthesitis was diagnosed by the attending pediatric rheumatologist in 45% of patients. Dactylitis was observed in 37% of patients and, when present, was typically evident early in the disease course: in 41 (80%) of 51 patients at the initial evaluation and in 47 (92%) of 51 patients within the first 6 months of disease. A histogram showing the age at onset of symptoms is provided in Figure 1. Two distributions are apparent, with one peaking at approximately 2 years of age and the other peaking in late childhood. Indeed, for 3567 the population as a whole the results of the KolmogorovSmirnov test showed clear evidence of non-normality (Lilliefors-corrected P ⬍ 0.001). Division into 2 subgroups according to age, optimally younger than age 5 years versus age 5 years and older, as determined by the assessment of multiple potential cutoff values, resulted in 2 populations, each deviating nonsignificantly from normal by the Lilliefors-corrected Kolmogorov-Smirnov test (P ⫽ 0.17 for children under age 5 years and P ⫽ 0.20 for children 5 years of age and older). Table 3 presents the clinical characteristics of these 2 subpopulations. The younger patients (n ⫽ 49) were significantly more likely to be female, to be ANA positive, and to have polyarticular disease at the time of presentation as well as at 6 months. Although the mean erythrocyte sedimentation rates (ESRs) and levels of C-reactive protein (CRP) were similar between subgroups, the mean platelet count was substantially elevated in younger patients, consistent with ongoing systemic inflammation. (The mean white blood cell count was also higher among younger patients, but this difference was expected as a result of the different mean ages of the subgroups; in contrast, the platelet count does not vary with age ). Older children were more likely to have axial disease and enthesitis. Uveitis occurred at a similar frequency in both groups and in most cases was asymptomatic and was discovered during routine ophthalmologic screening; one child in the older age group reported associated photophobia. Younger children were more likely to be treated with methotrexate, while treatment with sulfasalazine was far more common in the older children (Table 3). Despite receiving care that was relatively more aggressive, younger patients took substantially longer to enter remission (defined as the absence of clinically detectable synovitis or enthesitis, on or off medications) compared with their older counterparts (23 months versus 9.2 months; P ⫽ 0.044). Overall, our patients had an excellent functional outcome. At last evaluation, 81% of all patients were in Steinbrocker class I, and the remainder were in class II, with no differences between the agebased subgroups (13). Both subgroups were equally likely to be experiencing disease remission at the last recorded visit (60% of patients with disease onset before age 5 years versus 56% of those with disease onset at age 5 years or older [P ⫽ 0.66]; for remission off medications, 40% versus 41% [P ⫽ 0.92]). Although age is a very important guide for clinicians who are structuring a differential diagnosis, it has limited appeal as the defining characteristic of a complex clinical subgroup. To investigate whether age at symptom onset might serve as a marker of other clinical and 3568 STOLL ET AL Table 3. Clinical characteristics of patients according to age* Characteristic Female sex, % Duration of followup, median (IQR) months Psoriasis, % Age, median (IQR) years Joints affected, % Any axial Temporomandibular Shoulder Spine Sacroiliac Hip Any peripheral large joint Elbow Wrist Knee Ankle Any peripheral small joint Metacarpophalangeal Proximal interphalangeal Distal interphalangeal Metatarsophalangeal Dactylitis, % Enthesitis, % ESR, mean ⫾ SD mm/hour C-reactive protein, mean ⫾ SD mg/dl Platelet count, ⫻1,000 cells/l, mean ⫾ SD White blood cell count, ⫻1,000 cells/l, mean ⫾ SD ANA status, no. tested/no. positive (%) HLA–B27 status, no. tested/no. positive (%) Treatment, % Sulfasalazine Methotrexate Tumor necrosis factor inhibitors Any disease-modifying antirheumatic drug Course, % Oligoarticular at first visit Oligoarticular at 6 months† Oligoarticular course† Uveitis ever, no. positive/no. tested (%) Remission off therapy at last examination, %‡ Remission at last examination, %‡ Time to remission, median (IQR) months‡ Steinbrocker class I at last visit, %‡ ⬍5 years (n ⫽ 49) ⱖ5 years (n ⫽ 90) P 76 23 (9.2–46) 14 1.8 (1.4–2.5) 50 22 (7.5–52) 31 10 (7.7–12) 0.003 NS 0.029 ⬍0.001 10 4.1 4.1 0 0 4.1 82 14 24 61 59 76 18 26 14 16 63 22 27 ⫾ 17 0.8 ⫾ 1 408 ⫾ 98 10.5 ⫾ 4.8 25/39 (64) 1/3 (33.3) 26 7.8 2.2 3.3 1.1 14 79 12 26 60 47 47 18 23 6.7 11 22 57 25 ⫾ 21 0.5 ⫾ 0.8 337 ⫾ 96 7.8 ⫾ 2.6 29/78 (37) 5/10 (50) 0.031 NS NS NS NS NS NS NS NS NS NS 0.001 NS NS NS NS ⬍0.001 ⬍0.001 NS NS 0.001 0.002 0.006 NS 39 59 10 74 80 69 67 3/38 (7.9) 40 60 23 (11–49) 82 59 39 13 74 94 92 82 3/38 (7.9) 41 56 9.2 (3.4–37) 81 0.023 0.022 NS NS 0.007 0.002 NS NS NS NS 0.044 NS * P values less than 0.05 were considered significant. Bonferroni’s adjustment changed the significance threshold to P ⬍ 0.0013. See Table 1 for definitions. † Among those with at least 6 months of followup (for age younger than 5 years, n ⫽ 39; for age 5 years or older, n ⫽ 72). ‡ Reported for all patients with at least 2 visits (for age younger than 5 years, n ⫽ 45; for age 5 years or older, n ⫽ 84). laboratory features, we applied cluster analysis, using age at onset as a continuous variable and sex, ANA status, psoriasis, dactylitis, axial arthritis, enthesitis, and oligoarticular onset as categorical variables. The selection of these variables was informed by our earlier exploration of the age-based subgroups but was driven also by our assessment that they represent, to a substantial degree, the clinical heterogeneity observed in practice. Variables were chosen prior to the application of the cluster algorithm and were not further amended in the course of data analysis. For the purpose of cluster analysis, we included only patients for whom complete information was available. In practice, this led to the exclusion of 22 patients whose ANA status was not contained in the hospital medical record. Consistent with our earlier findings, cluster analysis detected 2 distinct populations within the data set. Allocation of patients to these clusters was highly robust, remaining stable to random scrambling of patient order in each of 100 iterations. The clinical characteristics of TWO POPULATIONS OF PATIENTS WITH JUVENILE PsA 3569 Table 4. Clinical characteristics of patients according to cluster assignment* Characteristic Female sex, % Duration of followup, median (IQR) months Psoriasis, % Age under 5 years, % Age, median (IQR) years Joints affected, % Any axial Temporomandibular Shoulder Spine Sacroiliac Hip Any peripheral large joint Elbow Wrist Knee Ankle Any peripheral small joint Metacarpophalangeal Proximal interphalangeal Distal interphalangeal Metatarsophalangeal Dactylitis, % Enthesitis, % ESR, mean ⫾ SD mm/hour C-reactive protein, mean ⫾ SD mg/dl Platelet count, ⫻1,000 cells/l, mean ⫾ SD White blood cell count, ⫻1,000 cells/l, mean ⫾ SD ANA status, no. tested/no. positive (%) HLA–B27 status, no. tested/no. positive (%) Treatment, % Sulfasalazine Methotrexate Tumor necrosis factor inhibitors Any disease-modifying antirheumatic drug Course, % Oligoarticular at first visit Oligoarticular at 6 months† Oligoarticular† Uveitis ever, no. tested/no. positive (%) Remission off therapy at last examination, %‡ Remission at last examination, %‡ Duration to remission, median (IQR) months‡ Steinbrocker class I at last visit, %‡ Cluster 1 (n ⫽ 40) Cluster 2 (n ⫽ 77) P 75 27 (10–49) 10 60 2.7 (1.4–8.5) 52 24 (8.8–55) 29 20 9.5 (6.2–11) 0.016 NS 0.022 ⬍0.001 ⬍0.001 12 5 2.5 0 0 5 75 22 30 55 58 100 38 42 12 20 100 35 26 ⫾ 16 0.7 ⫾ 1 394 ⫾ 106 9.6 ⫾ 3.5 28/47 (60) 0/2 (0) 25 7.8 3.9 3.9 1.3 13 84 9.1 27 64 48 32 12 18 7.8 7.8 1.3 52 24 ⫾ 22 0.5 ⫾ 1 340 ⫾ 92 8.1 ⫾ 3.8 26/77 (34) 5/9 (56) NS NS NS NS NS NS NS 0.045 NS NS NS ⬍0.001 0.001 0.005 NS NS ⬍0.001 NS NS NS 0.017 NS ⬍0.001 NS 52 62 18 82 75 67 58 2/31 (6.5) 26 54 25 (8.6–52) 82 56 43 12 75 97 95 89 3/37 (8.1) 47 59 15 (4.4–41) 82 NS 0.044 NS NS ⬍0.001 ⬍0.001 ⬍0.001 NS 0.026 NS NS NS * Variables used in cluster assignments are shown in boldface. P values less than 0.05 were considered significant. Bonferroni’s adjustment changed the significance threshold to P ⬍ 0.0013. See Table 1 for definitions. † Among those with at least 6 months of followup (for cluster 1, n ⫽ 35; for cluster 2, n ⫽ 62). ‡ Reported for all patients with at least 2 visits (for cluster 1, n ⫽ 41; for cluster 2, n ⫽ 71). the 2 clusters corresponded well with the features of our 2 groups defined by age alone (Table 4 and Figure 2). The median ages of children in the 2 clusters were 2.7 years (cluster 1) and 9.5 years (cluster 2), compared with 1.8 years and 10 years for the age-based grouping. Again, patients in the cluster of younger children were more likely to be female and to have small joint involvement, polyarticular disease, and ANA positivity, while patients in the older cluster were again equally divided among the sexes and were more likely to have oligoarticular disease. Using cluster analysis, differences in both axial joint involvement and enthesitis became nonsignificant, while the percentage of patients achieving remission off medications was significantly lower in the younger-age cluster. Most strikingly, the key defining feature of cluster assignment became the presence of dactylitis: 100% of patients in the younger-age cluster (cluster 1) 3570 STOLL ET AL Figure 2. Comparison of subpopulations of patients with juvenile psoriatic arthritis as defined by age at symptom onset (left) and cluster analysis (right). Antinuclear antibody (ANA) positivity is expressed as the percentage of patients for whom ANA data were available. had dactylitis, compared with only 1.3% of patients in the older-age cluster (cluster 2). Age at onset, although critical to formation of the clusters as determined by repetition of the analysis without this variable, was less clearly segregated: only 60% of patients within the dactylitis cluster had disease onset at an age younger than 5 years, while 20% of patients in the nondactylitis cluster also had disease onset at this age range. DISCUSSION In the absence of a biologic gold standard, the classification of rheumatologic diseases is obligatorily based on clinical, laboratory, and occasionally genetic data. For this purpose, close phenotyping is essential to ensure that subsequent research is not hampered by the conflation of distinct disorders. However, the relevant variables for such classification are not trivially ascertained. This is evident in the uncertainty surrounding the classification of adult PsA. Moll and Wright (18) categorized patients into 5 clinical subcategories based on patterns of joint involvement, yet these subcategories have had limited utility in clinical practice and in research, in part because joint involvement may evolve over time (19). This problem is compounded in pediatric rheumatology by the observation that children commonly present with arthritis years before the development of skin involvement (2,5,6,9,20). To the extent that such patients can be identified, for example via the Vancouver and ILAR criteria, observers have considered the possibility that important clinical subgroups of juvenile PsA may exist (11,21). Southwood et al (6) and Truckenbrodt and Hafner (5) observed a biphasic age-at-onset curve, while Shore and Ansell (2) and Roberton et al (9) noted that younger patients were more likely to be female. However, none of these series had the statistical power to discriminate clinical subpopulations. Using the Vancouver criteria, we assembled a series of patients with juvenile PsA; in our study, the number of patients was approximately twice the number of patients in each of the largest prior cohorts (2,9,11). Analysis of this data set provides the first statistical evidence that children with juvenile PsA fall into 2 distinct phenotypic subgroups. We observed that patients younger than age 5 years are more likely to be female and ANA positive and to have dactylitis; they are also more likely to experience polyarticular onset. Although the ESRs and the CRP levels at presentation were equivalent between subgroups, an elevated platelet count suggested more systemic inflammation in the younger children. In contrast, children older than age 5 years were evenly split between males and females and were more likely to be ANA negative, to exhibit enthesitis and axial joint involvement, and to present with and continue to have oligoarticular involvement. Perhaps surprisingly, uveitis was equally prevalent in both groups. In addition to these clinical differences, younger and older patients appeared to exhibit divergent responses to therapy. Sulfasalazine is rarely prescribed for children younger than age 2 years due to a paucity of safety data in this age group (22); therefore, it was not unexpected to observe that younger patients were treated less often with sulfasalazine and more often with methotrexate. Despite therapy with what is generally considered a more potent agent (23), disease in younger patients took almost twice as long to enter remission compared with disease in older children, although nearly 60% of children in both groups ultimately achieved disease control during the study period. These observations complement previous studies of HLA associations in juvenile PsA. Ansell and colleagues performed HLA typing on 70 children meeting the Vancouver criteria for juvenile PsA (11). Comparisons within the group suggested interesting HLA differ- TWO POPULATIONS OF PATIENTS WITH JUVENILE PsA ences between patients with onset before versus after 6 years of age. Indeed, it was recently proposed that age at onset may be quite important in the identification of etiologically homogeneous subtypes of juvenile arthritis (24). Early onset may reflect an aberrant response to an endemic infectious pathogen encountered within the first few years of life, while later onset could represent a response to accumulated injury or to an environmental trigger encountered more rarely or at a later phase in life. Indeed, age dependence is evident in both psoriasis and PsA in adults, with patients with disease onset at an age younger than 40 years exhibiting stronger HLA linkage and a higher genetic risk of relatives being similarly afflicted (25). Although analysis by age at onset confirmed indications from prior literature, we wished to refine our analysis to explore further the clinical and laboratory features that optimally identify subpopulations of patients with juvenile PsA. Cluster analysis is a statistical technique that can detect latent relationships within a complex data set between individuals with multiple distinct characteristics, in this case age at onset, sex, ANA status, oligoarticular involvement, dactylitis, enthesitis, axial disease, and psoriasis (16). Individuals are grouped together in clusters in an attempt to minimize differences between cluster members while maximizing differences between individuals in different clusters. Discretion in the choice of input variables and in the assumptions used to calculate similarity renders cluster analysis, as used here, an essentially exploratory statistical technique. Furthermore, the cluster analysis algorithms we used gave equal weights to differences in the various categorical variables, potentially distorting biologic relevance. For example, it may not be the case that 2 individuals who are ANA positive are “just as similar” as are 2 individuals of the same sex or 2 individuals who share the presence of dactylitis. With these caveats, cluster analysis provides an objective method to determine whether a heterogeneous group contains distinct subsets without assuming the primacy of any 1 variable (e.g., age at onset). Using cluster analysis, the optimum number of subpopulations in our cohort was found to be 2, with clinical and laboratory characteristics that corresponded well but not perfectly to those of the subgroups defined by age at onset alone (Table 4 and Figure 2). However, cluster analysis using the specified input variables generated the provocative alternative hypothesis that differences observed across the age spectrum result from the nonuniform age distribution of a subform of arthritis featuring dactylitis. This hypothesis would also be consistent with the age-dependent distribution of HLA 3571 subtypes observed by Ansell and colleagues (11) and invites speculation as to the manner in which the pathophysiology of dactylitis (tenosynovitis, small joint arthritis, periostitis, enthesitis) might reflect a distinct pathogenesis for this subform of disease (26,27). Our data are unable to adjudicate between the age-based subgroups and the dactylitis-based clusters, because both are consistent with all available information. Another alternate explanation for heterogeneity would be unmeasured factors, including perhaps most prominently the presence of HLA–B27 as a known risk factor for axial disease in adult PsA (28). In this series, the number of patients tested for the HLA–B27 antigen was low, likely reflecting the prevailing opinion of clinicians in our center that such testing rarely impacts on diagnostic or therapeutic decisions (29). Because previous studies have documented that the prevalence of HLA–B27 among patients with a diagnosis of juvenile PsA according to the Vancouver criteria is ⬍20%, it is clear that this factor could not, by itself, define either of the subgroups we identified (11,30). Published data for children meeting the Vancouver criteria for juvenile PsA suggest that HLA–B27 could be somewhat more common in children with an age at onset of ⬎6 years (3 [11.5%] of 26 versus 11 [25%] of 44; P ⫽ 0.17 [analyzed from raw data provided in ref. 11]). Correspondingly, of the 6 patients in our cohort who were positive for HLA–B27, 5 had disease onset at age 5 years or older; 3 of 6 HLA–B27–positive patients had axial disease compared with none of 7 patients who were negative for the antigen (P ⫽ 0.07, by Fisher’s exact test). To address the concern that the observed differences between subgroups reflect the presence or absence of HLA–B27, we repeated our analysis excluding every child with axial disease, thus maximally depleting HLA–B27–positive patients from the cohort. Despite excluding these 28 children, we still observed that younger children were more likely to be female, to demonstrate small joint involvement, to have a polyarticular course, and to have elevated platelet counts and ANA positivity, and were less likely to have enthesitis (all P ⬍ 0.05; data not shown). This analysis suggests that, even in the absence of axial disease and HLA–B27, older children have different epidemiologic and clinical features compared with their younger counterparts. Along with other findings in this essentially exploratory study, the role of HLA–B27 in juvenile PsA will require investigation and validation in other cohorts. The finding that children with PsA fall into 2 phenotypic groups is open to at least 2 interpretations. The first is that the arthritic manifestations of psoriasis change with the age of the patient at disease onset or with 3572 STOLL ET AL exposures for which age is a marker, such as infections, vaccinations, and environmental toxins. The second interpretation is that there are, in fact, 2 broad types of PsA in children, arising through the activity of distinct genetic and environmental factors. Based on the data presented here, we are unable to decide between these 2 possibilities. However, differences in sex ratios, ANA status, and HLA types (11) between the 2 phenotypic populations suggest that divergent pathophysiologic processes may be involved, thus favoring the second hypothesis. In summary, we identified 139 children who fulfilled the Vancouver criteria for juvenile PsA. Our data confirm the existence of 2 distinct subpopulations within this cohort: a group of younger children, most of whom are female, experience more small joint involvement and dactylitis and require longer to achieve disease remission, and an older group split evenly between boys and girls and with an increased incidence of axial disease and enthesitis. Alternately, roughly similar groups can be identified by the presence of dactylitis, a strong correlate of young age. Although both of these subgroups may still properly be considered to have arthritic manifestations of the psoriatic diathesis, they should no longer be assumed to be a homogeneous population for the purpose of etiologic and therapeutic studies. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. ACKNOWLEDGMENT We thank Dr. Bryce A. Binstadt for graphics expertise and for thoughtful review of the manuscript. 23. REFERENCES 24. 1. Lambert JR, Ansell BM, Stephenson E, Wright V. Psoriatic arthritis in childhood. Clin Rheum Dis 1976;2:339–52. 2. Shore A, Ansell BM. Juvenile psoriatic arthritis: an analysis of 60 cases. J Pediatr 1982;100:529–35. 3. Calabro JJ. Psoriatic arthritis in children. Arthritis Rheum 1977;20 Suppl 2:415–6. 4. Sills EM. Psoriatic arthritis in childhood. Johns Hopkins Med J 1980;146:49–53. 5. Truckenbrodt H, Hafner R. Psoriatic arthritis in childhood: a comparison with subgroups of chronic juvenile arthritis. Z Rheumatol 1990;49:88–94. 6. Southwood TR, Petty RE, Malleson PN, Delgado EA, Hunt DW, Wood B, et al. Psoriatic arthritis in children. Arthritis Rheum 1989;32:1007–13. 7. Petty RE, Southwood TR, Baum J, Bhettay E, Glass DN, Manners P, et al. Revision of the proposed classification criteria for juvenile idiopathic arthritis: Durban, 1997. J Rheumatol 1998;25:1991–4. 8. Petty RE, Southwood TR, Manners P, Baum J, Glass DN, Goldenberg J, et al. International League of Associations for 25. 26. 27. 28. 29. 30. Rheumatology classification of juvenile idiopathic arthritis: second revision, Edmonton, 2001. J Rheumatol 2004;31:390–2. Roberton DM, Cabral DA, Malleson PN, Petty RE. Juvenile psoriatic arthritis: followup and evaluation of diagnostic criteria. J Rheumatol 1996;23:166–70. Krumrey-Langkammerer M, Hafner R. Evaluation of the ILAR criteria for juvenile idiopathic arthritis. J Rheumatol 2001;28: 2544–7. Ansell B, Beeson M, Hall P, Bedford P, Woo P. HLA and juvenile psoriatic arthritis. Br J Rheumatol 1993;32:836–7. Southwood TR, Petty RE. Juvenile psoriatic arthritis. In: Cassidy JT, Petty RE, editors. Textbook of pediatric rheumatology. 5th ed. Philadelphia: Elsevier Saunders; 2005. p. 324–33. Steinbrocker O, Traeger CH, Batterman RC. Therapeutic criteria in rheumatoid arthritis. JAMA 1949;140:659–62. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990;1:43–6. Perneger TV. What’s wrong with Bonferroni adjustments. BMJ 1998;316:1236–8. Brian E, Landau S, Leese M. Cluster analysis. London: Edward Arnold; 2001. Marks PW. Data of importance in hematology. In: Handin RI, Lux SE, Stossel TP, editors. Blood: principles and practice of hematology. Philadelphia: Lippincott Williams & Wilkins; 2002. p. 2193, 2206. Moll JM, Wright V. Psoriatic arthritis. Semin Arthritis Rheum 1973;3:55–78. Helliwell PS, Taylor WJ. Classification and diagnostic criteria for psoriatic arthritis. Ann Rheum Dis 2005;64 Suppl 2:ii3–8. Wesolowska H. Clinical course of psoriatic arthropathy in children. Mater Med Pol 1985;17:185–7. Ansell BM. Juvenile psoriatic arthritis. Baillieres Clin Rheumatol 1994;8:317–32. Van Rossum MA, Fiselier TJ, Franssen MJ, Zwinderman AH, ten Cate R, van Suijlekom-Smit LW, et al, for the Dutch Juvenile Chronic Arthritis Study Group. Sulfasalazine in the treatment of juvenile chronic arthritis: a randomized, doubleblind, placebo-controlled, multicenter study. Arthritis Rheum 1998;41:808–16. Felson DT, Anderson JJ, Meenan RF. Use of short-term efficacy/ toxicity tradeoffs to select second-line drugs in rheumatoid arthritis: a meta-analysis of published clinical trials. Arthritis Rheum 1992;35:1117–25. Martini A. Are the number of joints involved or the presence of psoriasis still useful tools to identify homogeneous disease entities in juvenile idiopathic arthritis? J Rheumatol 2003;30: 1900–3. Rahman P, Elder JT. Genetic epidemiology of psoriasis and psoriatic arthritis. Ann Rheum Dis 2005;64 Suppl 2:ii37–9. Kane D, Greaney T, Bresnihan B, Gibney R, FitzGerald O. Ultrasonography in the diagnosis and management of psoriatic dactylitis. J Rheumatol 1999;26:1746–51. Olivieri I, Barozzi L, Favaro L, Pierro A, de Matteis M, Borghi C, et al. Dactylitis in patients with seronegative spondylarthropathy: assessment by ultrasonography and magnetic resonance imaging. Arthritis Rheum 1996;39:1524–8. Marsal S, Armadans-Gil L, Martinez M, Gallardo D, Ribera A, Lience E. Clinical, radiographic and HLA associations as markers for different patterns of psoriatic arthritis. Rheumatology (Oxford) 1999;38:332–7. Gran JT, Husby G. HLA-B27 and spondyloarthropathy: value for early diagnosis? J Med Genet 1995;32:497–501. Hamilton ML, Gladman DD, Shore A, Laxer RM, Silverman ED. Juvenile psoriatic arthritis and HLA antigens. Ann Rheum Dis 1990;49:694–7.