Trajectories of Physical Function in Canadian Children With Juvenile Idiopathic Arthritis

Abstract

Objective

We describe trajectories of physical function in children newly diagnosed with juvenile idiopathic arthritis (JIA) and identify trajectories with persisting functional impairments and associated baseline characteristics.

Methods

We included patients enrolled in the Canadian Alliance of Pediatric Rheumatology Investigators (CAPRI) Registry between 2017 and 2024 whose parents provided Kids Disability Screen (KDS) scores from at least two visits (from 0 = no disability to 10 = severe disability). Analyses included descriptive statistics, locally weighted scatterplot smoothing, Kaplan‐Meier plots of time to KDS = 0, latent class trajectory analysis (LCTA), and classification trees to predict trajectories with persisting impairments.

Results

We included 940 patients providing 7,351 KDS scores starting at a median 6 days after diagnosis up to 7 years. Baseline KDS scores were highest for rheumatoid factor–positive polyarthritis (mean 4.9, SD 3.1) and lowest for psoriatic arthritis (mean 2.4, SD 2.5), and mean scores improved within the first year in all JIA categories. Median time to KDS = 0 was 36 weeks (from 13 weeks for systemic arthritis to 73 weeks for undifferentiated arthritis). LCTA identified five trajectories: one with little measurable impairment (38% of patients), two with impairments resolving in one to two years (24%), and two with persisting mild impairments (38%). A baseline KDS ≥1.5 had a sensitivity of 0.79 and specificity of 0.48 to detect persisting mild impairment trajectories. There were no trajectories with persisting moderate or severe impairments.

Conclusion

Most children with JIA in this cohort had mild to moderate functional impairments at diagnosis that resolved in one to two years, but one in three followed trajectories with mild persisting impairments.

INTRODUCTION

Juvenile idiopathic arthritis (JIA) is the most common chronic rheumatic disease of childhood, with reported worldwide prevalence ranging from 4 to 400 cases per 100,000 children. ¹ Children and adolescents with JIA often experience pain, joint swelling, and functional impairments. Functional impairments can take many forms, including difficulties engaging in play or sports and in daily activities, such as getting dressed or writing. JIA is classified into seven categories according to the International League of Associations for Rheumatology (ILAR) criteria: oligoarthritis, rheumatoid factor (RF)–negative polyarthritis, RF‐positive polyarthritis, and systemic, psoriatic, enthesitis‐related, and undifferentiated arthritis. ² JIA category significantly impacts a child's expected disease course and medical treatment. ³

SIGNIFICANCE & INNOVATIONS.

• In this Canadian longitudinal study of 940 children newly diagnosed with juvenile idiopathic arthritis (JIA) between 2017 and 2024, most had mild to moderate functional impairments at the time of diagnosis that improved quickly with a median time to optimal function of 36 weeks.

• At diagnosis, rheumatoid factor–positive polyarthritis had the largest functional impact, and psoriatic arthritis had the least, and patients with systemic arthritis and oligoarthritis attained optimal function earlier than patients in other JIA categories.

• In latent class trajectory analysis, 38% of patients were assigned to trajectories with persisting mild functional impairments, and a baseline Kids Disability Screen ≥1.5 was able to identify 79% of these patients.

• There were no functional trajectories with persisting moderate or severe impairments, but 8% of patients reported moderate to severe impairment at least once beyond two years after diagnosis.

Early studies have reported that juvenile arthritis was frequently associated with lifelong disability. In a 1981 study, about half of patients followed for 10 to 22 years became handicapped or experienced some functional limitation, and 11.3% had severe limitations. ⁴ With modern targeted treatment with conventional synthetic, targeted synthetic, and biologic disease‐modifying antirheumatic drugs (DMARDs), it is anticipated that most children with JIA will quickly improve in function.

In Canada, achievement of clinically inactive disease within 70 weeks of JIA diagnosis has increased from 64% to 83% when comparing 2005 to 2010 and 2017 to 2021 inception cohorts. ⁵ Despite these improvements in control of disease activity, it is known that some children still experience substantial functional impairments, but the characteristics of this patient population have not been systematically described. ⁶ There is limited prospective data on physical function trajectories in patients on modern treatments, including how often impairments persist and which baseline factors predict them. A systematic review found that the few proposed predictors were not confirmed in other studies, ⁷ and there is a dearth of information on long‐term function of patients with disease onset in the biologic era. ⁸

The World Health Organization (WHO) International Classification of Functioning, Disability, and Health (ICF) is the international framework for measuring health and disability in individuals and broader populations. ⁹ In ICF, functioning refers to all body functions, activities, and participation, whereas disability refers to all impairments (significant deviations or loss of body functions), activity limitations, and participation restrictions. ⁹ Worse functional impairments are associated with poorer health‐related quality of life, ¹⁰ , ¹¹ and better physical functioning is consistently associated with better overall pediatric quality of life scores. ¹² In children and adolescents ages 8 to 16 years, a key period for development, functional impairments can cause delays in the optimization of gross motor skills and disrupt emotional and social development, ultimately affecting overall quality of life. ¹²

In Canada, most children with JIA receive services to address function, such as physiotherapy and/or occupational therapy assessment and management. These interventions aim to reduce pain, improve range of motion in joints, increase strength, and improve participation in daily activities. ¹³ , ¹⁴ By introducing these services closer to disease onset, patients may be more likely to improve physical function and well‐being. Implementing preventive measures rather than treating established deficits is thought to have long‐term positive effects on function.

This study aimed to describe modern longitudinal trajectories of physical function in children with different JIA categories and to identify data‐driven trajectories of physical function that may lead to persisting functional impairments and the baseline characteristics that may predict these trajectories. Early identification of patients at high risk for long‐term functional impairments would enable timely, targeted preventive interventions.

METHODS

Patients

Data collected between February 2017 and April 2024 from 20 academic and community pediatric rheumatology clinics across Canada participating in the Canadian Alliance of Pediatric Rheumatology Investigators (CAPRI) Registry were analyzed. (Members of the CAPRI Registry Investigators are included in Acknowledgements). The CAPRI Registry prospectively collects information on disease course, treatment, and outcomes among Canadian children diagnosed with JIA. ¹⁵ Children and adolescents with disease onset before the age of 16 years are eligible to be enrolled within the first three months after their JIA diagnosis, if the family is able to complete questionnaires in English or French. Demographic and clinical characteristics of patients are reported by parents and physicians at enrollment and at every clinic visit until discharge from clinic or transfer to adult care. The CAPRI Registry study was approved by research ethics boards at all participating centers. Informed consent and appropriate assent were obtained from parents and participants, respectively. Central approval was provided by The University of British Columbia Children's & Women's Health Centre of British Columbia Research Ethics Board (H16‐00309). For this study, registry participants whose parents had completed at least two functional assessment questionnaires by April 2024 were included; patients with missing recorded date of diagnosis or JIA category were excluded.

Measures

Physical function, our main outcome, was assessed using the Kids Disability Screen (KDS). ¹⁶ KDS is the mean of numerical responses (from 0 to 10 in a 21‐point scale) to two questions: “Does your child usually need help from you or another person because of arthritis?” and “Is it hard for your child to run and play because of arthritis?” The KDS score ranges from 0 = no disability to 10 = severe disability and has psychometric properties comparable to the 32‐item Childhood Health Assessment Questionnaire (CHAQ) disability index. ¹⁶ Based on their relationship with CHAQ scores, KDS scores >0 to 1.5 indicate mild disability, >1.5 to 4 mild to moderate disability, and >4 moderate to severe disability. In a validation cohort, the KDS had a standardized response mean of 0.67 compared with 0.62 for CHAQ, and it detected functional impairments in 44% of children with no impairment detected by CHAQ (CHAQ = 0). ¹⁶ The answers to the two questions can also be combined with other patient characteristics to calculate an estimated CHAQ (eCHAQ) score. Concordance between eCHAQ and CHAQ was very good with a weighted kappa of 0.78. ¹⁶ The parents of all registry participants were asked to complete the KDS at enrollment and at each subsequent clinic visit. At enrollment (and annually thereafter) parents were also asked to complete the CHAQ.

Selected variables collected at enrollment in the registry were examined as possible predictors of physical function trajectories. These included child's age at diagnosis, sex, disease duration (length of time between the first symptom and diagnosis), number of active joints at enrollment, physician global assessment of disease activity (PGA; 0–10), acute phase reactants (highest erythrocyte sedimentation rate and highest C‐reactive protein since onset), and pain intensity (0–10). Additionally, we examined baseline domain scores of the Juvenile Arthritis Quality of Life Questionnaire (JAQQ; gross motor function, fine motor function, psychosocial function, and systemic symptoms) ¹⁷ and scores for fatigue and pain interference using Patient‐Reported Outcome Measurement System (PROMIS) short parent‐proxy questionnaires. ¹⁸ , ¹⁹

Analysis

Baseline characteristics of the cohort were described with numbers and percentages or medians and 25th and 75th percentiles, as appropriate for categorical, ordinal, and continuous variables. Trajectories of mean KDS scores for each JIA category following diagnosis were plotted with locally weighted scatterplot smoothing (LOWESS) with a bandwidth of 0.8 using STATA 15 software (StataCorp LLC). Kaplan–Meier survival analysis was used to estimate and display median time for children to attain optimal functioning (KDS = 0), for the overall cohort and by JIA category. Pairwise comparisons between JIA categories were done with log‐rank tests with Bonferroni correction for multiple comparisons. As there were 21 comparisons, the Bonferroni threshold was P < 0.002. We calculated the frequency of visits with KDS scores >4 points after 2 years post diagnosis and examined their relationship with disease flares. Flare was defined as an increase of one point or more in the PGA and starting or increasing antirheumatic medications.

Latent class trajectory analysis (LCTA) was used to identify groups of patients with similar trajectories of function irrespective of their JIA category, ²⁰ starting at the time of diagnosis and for up to four years. This was done in a subset of participants who had at least four visits recorded in the registry and at least two valid KDS scores. After multiple imputation of missing scores and exclusion of the baseline score (because we were interested in examining the predictive value of baseline KDS scores), we explored LCTA models from one to six trajectories and selected the model with the best values of both the Akaike Information Criterion and the Bayesian Information Criterion. ²¹ In secondary analyses, we also explored LCTA using eCHAQ scores. As age and sex are required for eCHAQ calculation, patients with missing recorded data on age or sex were excluded from this analysis.

Recursive partitioning was used to identify simple decision trees that could be used in clinic to detect patients assigned to latent classes with persisting impairments. ²² Multiple imputation by chained equations was used to impute values in 5 data sets for the LCTA with averages taken as per Rubin's rules ²³ and in 20 data sets for the recursive partitioning analysis. Our recursive partitioning (decision tree) analyses focused on predicting patients who develop persisting impairments versus those with resolving or no impairments. Because our aim was to identify a simple decision tree that could be used at baseline to detect most patients likely to follow trajectories with persisting impairments, we focused on sensitivity calculations and assessed scenarios when a KDS was available, a CHAQ was available, or no patient reported measures were available. Because we had 20 imputed data sets and there are not established methods to combine 20 trees that may select different variables and cutoff values, we combined all 20 data sets first and then constructed 1 tree that was agreed to by the bulk of the data, with imputation uncertainty reflected by allowing proportional node assignments. We used the full stacked data set with 20 rows of imputed data per patient to build trees containing one to five splits (two to six terminal nodes). To objectively evaluate the predictive ability of each different tree size, we performed five‐fold cross‐validation on the stacked data, fitting classification trees to each of the five folds. The folds were created by randomly assigning each patient to a fold, ensuring that all 20 observations per patient were assigned to the same fold. Missing data imputation, LCTA, and recursive partitioning computations were done using R software packages “mice,” “lcmm,” and “rpart,” respectively, ²⁴ and they are described in detail in the online Supplementary Materials.

RESULTS

A total of 1,125 patients had been enrolled in the CAPRI Registry within three months of diagnosis by April 2024. After excluding patients with missing JIA category (n = 10) and those who had less than two KDS scores (n = 175), a total of 940 patients were included in our study. Among 8,850 clinic visits recorded in the registry for these patients, there were 7,351 KDS scores available (83% of visits), 6,088 of them within three years of JIA diagnosis. Participant's median follow‐up time was 787 days after diagnosis (25th, 75th centiles 383–1,350), and 333 patients (35.4%) were followed for more than three years. The median number of visits per patient was 5 (25th, 75th centiles 2–8) with a maximum of 37 visits. Patients included in the study were comparable to excluded patients (Table 1), except that excluded patients were more recently diagnosed and followed for a shorter period, a median of 70 days following diagnosis (25th, 75th centiles 0–265).

Table 1.

Baseline characteristics of the cohort at enrollment^*

Characteristic	Included patients	Excluded patients
Number of patients	940	185
Age, y	9.1 (4.3, 13.2)	11.2 (7.0, 14.0)
Female, n (%)	587 (63.2)	101 (62.0)
Time since diagnosis, d	6 (0, 52)	0 (0, 42)
Time since disease onset, d	163 (90, 307)	152 (90, 381)
Follow‐up time, d	787 (383, 1350)	70 (0, 265)
Visits per patient	5 (2, 8)	2 (1, 5)
JIA category, a n (%)
Oligoarthritis	426 (45.3)	84 (46.9)
Polyarthritis RF‐negative	181 (19.3)	30 (16.8)
Enthesitis‐related arthritis	135 (14.4)	30 (16.8)
Psoriatic arthritis	52 (5.5)	14 (7.8)
Systemic arthritis	46 (4.9)	8 (4.5)
Polyarthritis RF‐positive	37 (3.9)	2 (1.1)
Undifferentiated	63 (6.7)	11 (6.1)
Active joint count	2 (1, 6)	2 (1, 4)
Physician global assessment	4.0 (2.5, 6.0)	3.0 (1.5, 5.0)
Parent global assessment	2.0 (0.5, 5.0)	2.0 (1.0, 5.0)
Pain in last week	3.0 (1.0, 6.0)	3.0 (1.0, 6.0)
CHAQ disability index	0.375 (0.0, 1.0)	0.375 (0.0, 1.0)
KDS	2.50 (0.50, 5.25)	2.625 (0.5, 5.5)
Whole cohort, mean (SD)	3.1 (2.9)	3.3 (2.9)
Oligoarthritis, mean (SD)	2.7 (2.9)	2.8 (2.7)
Polyarthritis RF‐negative, mean (SD)	3.6 (2.9)	3.9 (3.2)
Enthesitis‐related arthritis, mean (SD)	3.6 (2.6)	3.4 (2.6)
Psoriatic arthritis, mean (SD)	2.4 (2.5)	4.5 (3.1)
Systemic arthritis, mean (SD)	2.7 (3.3)	2.8 (3.3)
Polyarthritis RF‐positive, mean (SD)	4.9 (3.1)	1.9 (1.6)
Undifferentiated, mean (SD)	3.2 (2.7)	4.5 (3.1)

^*Numbers are median (25th, 75th centiles) unless otherwise specified. CHAQ, Childhood Health Assessment Questionnaire; JIA, juvenile idiopathic arthritis; KDS, Kids Disability Screen; RF, rheumatoid factor.

^a JIA category assigned by physician at diagnosis.

The demographic and clinical characteristics of participants at enrollment are summarized in Table 1. Participants were enrolled a median of six days after diagnosis (25th centile 0, 75th centile 52) at a median age of 9.1 years (25th centile 4.3, 75th centile 13.2), and 63.2% were female. Most patients had oligoarthritis (45.3%), RF‐negative polyarthritis (19.3%), or enthesitis‐related arthritis (ERA; 14.4%). The median KDS score upon enrollment was 2.50 (25th centile 0.50, 75th centile 5.25), and the median CHAQ was 0.375 (25th centile 0, 75th centile 1.0). The mean KDS score at enrollment was 3.1 (SD 2.9), with children with RF‐positive polyarthritis having the highest mean KDS at enrollment (4.9, SD 3.1) and children with psoriatic arthritis having the lowest (2.4, SD 2.5) (Table 1).

Trajectories of physical function by JIA category

Figure 1 shows trajectories of mean KDS scores for each JIA category obtained with LOWESS and the dispersion in KDS scores. Trajectories varied across JIA categories. As stated in the previous paragraph, patients with RF‐positive polyarthritis reported the highest KDS scores shortly after diagnosis, and patients with psoriatic arthritis reported the lowest scores. Mean KDS scores improved for all categories within the first year after diagnosis and then stabilized, with patients with ERA having somewhat higher mean KDS scores than the other categories up to four years. Because the number of patients followed for more than four years for infrequent JIA categories was small, LOWESS trajectories in Figure 1 should be interpreted with caution beyond this time. Although mean trajectories improved in the first year, 75 patients (8.0%) reported KDS >4 at least once beyond the initial two years after diagnosis (Figure 1B). There were 148 study visits with a KDS score >4 after two years of diagnosis in these 75 patients; 31 visits (20.9%) had a concomitant disease flare, 107 visits (72.3%) did not, and 10 visits (6.8%) did not have enough information to determine the presence of a flare. The median PGA value for these visits was 1 (25th, 75th centiles 0–3), with a KDS >4 reported in 50 visits with a PGA of 0.

Figure 1.

Trajectories of physical function by JIA category. Shown are (A) the trajectories of physical function by JIA category following diagnosis and (B) KDS score dispersion for the whole cohort. Trajectories were plotted using LOWESS with a bandwidth of 0.8. ERA, enthesitis‐related arthritis; JIA, juvenile idiopathic arthritis; KDS, Kids Disability Screen; LOWESS, locally weighted scatterplot smoothing; RF, rheumatoid factor.

Time to attain optimal function

A Kaplan–Meier survival analysis of the rate at which patients in different JIA categories attain an optimal KDS score of 0 showed substantial differences across JIA categories (Figure 2). The median time for the overall cohort to attain KDS = 0 was 36 weeks after diagnosis (25th, 75th centiles 15, 96). Patients with systemic arthritis attained a KDS score of 0 sooner, at a median of 13 weeks (25th, 75th centiles 7, 40), followed by patients with oligoarthritis at a median of 26 weeks (25th, 75th centiles 12, 61). Patients with undifferentiated arthritis took the longest at a median of 73 weeks (25th, 75th centiles 24, 104). Pairwise comparisons across JIA categories with log‐rank tests showed that patients with systemic arthritis and oligoarthritis had significantly faster attainment of a KDS of 0, relative to patients with polyarthritis RF‐negative and ERA (P < 0.001; Figure 2).

Figure 2.

Time to first attainment of an optimal KDS score of 0. Shown is (A) the probability of attaining an optimal KDS score of 0 over 3 years after JIA diagnosis by JIA category. Shown is (B) the median (25th, 75th centiles) number of weeks to attain a KDS score of 0 for each JIA category and for the overall cohort. Shown is (C) P values for pairwise comparisons of JIA categories using the log‐rank test. Comparisons with P < 0.05 are bolded. Considering the multiple pairwise comparisons in (C) (n = 21), values of P < 0.002 would correspond to the Bonferroni correction threshold. ERA, enthesitis‐related arthritis; JIA, juvenile idiopathic arthritis; KDS, Kids Disability Screen; RF, rheumatoid factor.

LCTA

LCTA included 727 patients with at least four recorded visits and two valid KDS scores. These patients attended 7,211 visits. The best fitting LCTA model identified five distinct data‐driven trajectories of physical function (Figure 3). There was one trajectory with little measurable functional impairment throughout four years of follow‐up (class 5, 38% of patients), two trajectories with initial impairments that resolved during follow‐up (class 1 with moderate impairment in 5% of patients and class 2 with mild to moderate impairment in 19% of patients), and two trajectories associated with persisting mild functional impairments (class 3 with initial mild to moderate impairment that improves but does not resolve in 24% of patients and class 4 with persisting mild impairment throughout in 14% of patients). There were no trajectories with persisting moderate or severe impairments identified by LCTA.

Figure 3.

Predicted KDS trajectories of physical function by LCTA with 95% CIs. Shown are the data‐driven trajectories of KDS scores over the 4 years after diagnosis, modeled using LCTA, and the proportion of patients predicted to follow each trajectory. CI, confidence interval; JIA, juvenile idiopathic arthritis; KDS, Kids Disability Screen; LCTA, latent class trajectory analysis.

LCTA looks at patterns within the data suggesting patient subgroups that share similar trajectories; then it calculates probabilities for each patient to belong to each of the identified classes/trajectories. The online Supplementary Materials report the diagnostic measures that led to choosing our five‐trajectory model. We assigned each patient to the class with the highest probability. The mean posterior classification probabilities for patients to be assigned to classes 1 to 5 were 0.66, 0.62, 0.64, 0.71, and 0.82, with 42%, 34%, 32%, 50%, and 68% of assigned patients having a probability above 0.70, respectively. The best fitting LCTA model when using eCHAQ scores, instead of KDS scores, identified three trajectories, and none of them had persisting functional impairments (see Supplementary Materials).

Many baseline indicators of disease activity, symptoms, function, and quality of life were significantly different across the five physical function trajectories (Table 2). The proportion of patients with oligoarthritis ranged from 15.8% in patients with initial moderate impairment (class 1) to 53% in patients with little measurable impairment in class 5. The median number of affected joints was 5 (25th, 75th centiles 2, 10) in class 1, compared to 2 (25th, 75th centiles 1, 5) in class 5 (P < 0.0001 across trajectories). The median pain score was 4.5 in class 1 (25th, 75th centiles 3, 7.5), compared to 1.5 in class 5 (IQR 0, 3.5; P < 0.001). Other parent‐reported outcomes followed a similar pattern with greater severity in class 1 and lesser severity in class 5, although class 3 had the highest levels of pain (median 5; 25th, 75th centiles 3, 7.5) and parent global assessment score (median 4; 25th, 75th centiles 2, 6). Joints important for ambulation, such as hip, foot, and ankle, showed a similar gradient, with more frequent involvement in class 1 (18.4% with hip and 60.5% with ankle involvement) and lesser involvement in class 5 (5.8% and 37.7%, respectively). In summary, the milder the disease presentation at baseline, the more likely patients would follow a trajectory with little measurable impairment.

Table 2.

Baseline characteristics of patients in each of five physical function trajectories identified with latent class trajectory analysis^*

Characteristic	Class 1: initial moderate impairment	Class 2: initial mild to moderate impairment	Class 3: initial mild‐moderate, persisting mild	Class 4: persisting mild impairment	Class 5: little measurable impairment	P value
Patients, n (%)	38 (5)	137 (19)	176 (24)	100 (14)	276 (38)
Age, y	8.4 (3.5, 11.8)	8.6 (3.8, 12.8)	10.7 (5.2, 13.8)	9.5 (4.2, 13.1)	7.1 (3.2, 12.0)	<0.001
Female, n (%)	23 (60.5)	84 (61.3)	120 (68.2)	64 (64.0)	171 (62.0)	0.665
Time to diagnosis, d	151 (73, 347)	104 (58, 213)	160 (78, 364)	139 (63, 364)	118 (61, 236)	0.049
JIA category, a n (%)						0.001
Oligoarthritis	6 (15.8)	47 (34.3)	72 (40.9)	50 (50)	148 (53.6)
Polyarthritis RF‐negative	11 (28.9)	34 (24.8)	35 (19.9)	16 (16)	49 (17.7)
Enthesitis‐related arthritis	5 (13.2)	23 (16.8)	31 (17.6)	13 (13)	22 (8.0)
Psoriatic arthritis	3 (7.9)	11 (8.0)	9 (5.1)	3 (3.0)	15 (5.4)
Systemic arthritis	2 (5.3)	6 (4.4)	8 (4.6)	6 (6.0)	18 (6.5)
Polyarthritis RF‐positive	2 (5.3)	7 (5.1)	9 (5.1)	6 (6.0)	9 (3.3)
Undifferentiated	9 (23.7)	9 (6.6)	12 (6.8)	6 (6.0)	15 (5.4)
Active joint count	5 (2, 10)	3 (2, 8)	3 (1, 6)	2 (1, 5)	2 (1, 5)	<0.001
Limited joint count	3 (1, 7)	3 (1, 5)	2 (1, 4)	1 (1, 3)	1 (1, 3)	<0.001
Hip involvement, n (%)	7 (18.4)	20 (14.6)	27 (15.3)	14 (14.0)	16 (5.8)	0.005
Foot or ankle involvement, n (%)	23 (60.5)	83 (60.6)	87 (49.4)	34 (34.0)	104 (37.7)	<0.001
Wrist involvement, n (%)	14 (36.8)	46 (33.6)	60 (34.1)	27 (27.0)	74 (26.8)	0.313
Presence of enthesitis, n (%)	5 (14.3)	14 (10.8)	22 (13.6)	9 (9.0)	10 (4.2)	0.013
Physician global assessment	4.2 (3, 7)	5 (3, 7)	4 (2.5, 6)	3.5 (2, 5)	3 (2, 5)	<0.001
Parent global assessment	3.7 (1.5, 7)	2.5 (1, 5)	4 (2, 6)	2 (0.5, 4)	0.5 (0, 2)	<0.001
Pain in last week (parent)	4.5 (3, 7.5)	3 (2, 7)	5 (3, 7.5)	3 (1.5, 5)	1.5 (0, 3.5)	<0.001
CHAQ disability index	1.1 (0.5, 1.2)	0.5 (0.1, 1.0)	0.7 (0.2, 1.4)	0.3 (0.1, 0.9)	0.1 (0, 0.5)	<0.001
KDS	5.6 (3.5, 7.7)	3.5 (1.2, 6.0)	5 (2.5, 7.0)	2 (0.7, 4.0)	0.5 (0, 2.5)	<0.001
ESR, mm/h	24 (14, 44)	25 (12, 41)	20 (9, 42)	18 (10, 44)	16 (7, 37)	0.062
CRP, mg/L	9 (4, 26)	11 (3, 31)	6 (1, 14)	5 (2, 15)	4 (1, 17)	<0.001
Pain interference score	70 (65, 74)	66 (60, 71)	69 (64, 74)	65 (56, 71)	57 (46, 65)	<0.001
Fatigue score	64 (60, 69)	60 (51, 67)	64 (58, 69)	60 (44, 66)	52 (43, 60)	<0.001
JAQQ total score	3.8 (2.9, 4.7)	3.0 (2.2, 4.4)	3.6 (2.7, 4.8)	2.7 (1.9, 3.7)	1.8 (1.4, 2.8)	<0.001

^*Values are median (25th, 75th centiles) unless otherwise specified. P value calculated across the five trajectories with Kruskal–Wallis test or chi‐square test, as appropriate. CHAQ, Childhood Health Assessment Questionnaire; CRP, C‐reactive protein; ESR, erythrocyte sedimentation rate; JAQQ, Juvenile Arthritis Quality of Life Questionnaire; JIA, juvenile idiopathic arthritis; KDS, Kids Disability Screen; RF, rheumatoid factor.

^a JIA category assigned by the pediatric rheumatologist at diagnosis.

Decision trees

When we analyzed scenarios in which baseline KDS scores were available, recursive partitioning selected KDS as the most discriminating first variable, and a cutoff of KDS ≥1.5 detected 79% of patients in trajectories with persisting mild deficits (ie, sensitivity = 0.79), with a specificity of 0.48 (Figure 4). When we analyzed scenarios in which a baseline CHAQ (but not a KDS) was available, recursive partitioning selected a CHAQ score ≥0.5 as the most discriminating variable cutoff with sensitivity of 0.59 and specificity of 0.59. Adding additional baseline variables to decision tress that included KDS or CHAQ scores did not substantially improve detection of patients with persisting mild functional impairments. In scenarios in which no baseline patient‐reported measures were available, age was the most discriminating variable and a cutoff of ≥ 9.5 years had a sensitivity of 0.54 and specificity of 0.60 for detecting patients with persisting mild functional impairments.

Figure 4.

Results of decision tree analyses. Shown is (A) the example of a 6‐node decision tree when a KDS is available at baseline and (B) the variables and cut‐offs that optimize detection of patients with trajectories of persisting mild impairments under different circumstances. ACTIVE_J, number of active joints; CHAQ, Childhood Health Assessment Questionnaire; KDS, Kids Disability Screen; LIMITED_J, number of joints with limited movement; ND, no persisting disability; PAIN_PAR, pain intensity in the last week reported by parent; PARENT_G, parent global assessment of well‐being; PD, persisting disability; PROS, patient‐reported outcomes. Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/doi/10.1002/acr.80039/abstract.

DISCUSSION

This study examined modern trajectories of physical function in a large Canadian JIA inception cohort with full representation of all JIA categories. We found that, shortly after JIA diagnosis, functional impairments measured with the KDS varied across JIA categories. Initial KDS scores were worse for patients with RF‐positive polyarthritis and best for psoriatic arthritis. The mean KDS scores decreased during the first year after diagnosis for all JIA categories and then stabilized. A minority of children in each category reported significant impairments at visits beyond two years after diagnosis, and only one fifth of these visits were associated with documented arthritis flares. Children with systemic arthritis and oligoarthritis attained an optimal KDS score of 0 earlier than those in other JIA categories.

LCTA found five data‐driven physical function trajectories with one in three children following trajectories with persisting mild functional impairments. There were no trajectories with persisting moderate or severe functional impairments. Because patients were recruited between 2017 and 2024 and the length of follow‐up varied with most patients having under four years of follow‐up, the LCTA trajectories are presented for the first four years after diagnosis. Although multiple baseline characteristics differed across the five trajectories, in our decision tree analyses, an initial KDS ≥1.5 performed best and identified 79% of patients with persisting mild impairments. That said, our assignment of patients to the LCTA trajectory with greatest probability should be interpreted with caution because often the posterior probability of assignment was less than ideal (<70%).

Although, to our knowledge, this is the first study to examine modern trajectories of physical function in children with JIA, multiple studies have examined trajectories of other disease outcomes, including disease activity, ²⁵ , ²⁶ pain, ²⁷ and health‐related quality of life. ²⁸ Shoop‐Worrall et al reported trajectories of active joint counts, PGA, and parent global assessment of well‐being, identifying six trajectory clusters. ²⁵ Balay‐Dustrude et al assessed predictors of disease activity and found correlations across polyarticular RF‐negative, psoriatic, and enthesitis‐related JIA and worse disease activity at 12 and 24 months. Active disease at 3 months post diagnosis and older age were also associated with higher disease activity at 12 and 24 months. ²⁶

Treatments were not specifically examined in our study; however, therapies used in the CAPRI Registry have been recently described, highlighting a step‐up approach with escalation of treatment when there is an inadequate clinical response to prior therapy, consistent with current practice guidelines. ⁵ , ²⁹ Practices for discontinuing DMARDs in patients with inactive disease vary, although children with polyarticular arthritis typically remain on DMARDS for a longer duration. ³

Our cohort, with patients newly diagnosed with JIA between 2017 and 2024, more accurately reflects current functional trajectories, relative to studies in patients diagnosed before the widespread use of biologic DMARDs. These treatments have made significant impacts on the expectations of remission in children with JIA. ³⁰ Oliveira Ramos et al highlighted that long‐term longitudinal studies of JIA cohorts with disease onset in the biologic era are a necessary next step in JIA research. ⁶ Joint damage is considered an important factor leading to physical disability, and with full control of disease activity with modern medications, it is expected that joint damage is mitigated or prevented. ⁶ , ³¹

The main strengths of this study are the large cohort of children with JIA that is representative of all JIA categories and the data collection at every clinic visit using a validated brief measure of physical function resulting in thousands of assessments. Other studies of function in JIA have studied “convenience” populations, in which they collect information from all patients who attend the clinic within a certain time frame. This may lead to overrepresentation of children with functional impairments because patients who are clinically worse tend to have more frequent visits and, thus, a higher likelihood of being included in those studies. Collecting data from families across many locations in a country with a universal health care system and well‐developed pediatric rheumatology services resulted in a demographic that closely represents the general population of children with JIA. Collecting data at every visit to the clinic provides closer monitoring and more accurate detail about the longitudinal trajectories of function than could be provided by periodic questionnaires completed once or twice a year.

Our study should be interpreted considering several limitations. First, 17% of recorded visits to clinic for the included patients had missing KDS scores. Although questionnaires are often missing simply because families forget to complete them, there may be an association between having inactive disease and not completing questionnaires. Families often report that they feel questionnaires are not necessary when they have “nothing to report.” In our study, missing data were addressed with multiple imputation before LCTA. That said, we acknowledge that other factors may influence KDS scores and there are limitations to the accuracy of imputation of patient‐reported measures.

Second, there may be understandable hesitancy to directly compare results using the KDS to those obtained with the CHAQ, given the latter's widespread use and rigorously verified metrics in multiple settings and languages. The KDS used in this study has been shown to be highly sensitive at detecting the presence of functional impairments in children with JIA and may have a lesser ceiling effect and greater responsiveness than the CHAQ disability index in this population. ¹⁶ Interestingly, the use of the eCHAQ scores in LCTA in our cohort failed to detect the trajectories with mild persisting impairments detected using KDS scores. Being a short index with only two items makes it feasible to collect KDS scores at every clinic visit, relative to the 30+ item CHAQ. When designing the CAPRI Registry, it was felt that sustaining ongoing reliable completion of the CHAQ questionnaire at every visit to the clinic across all centers was not feasible, and feasibility was also the reason why Canadian pediatric rheumatologists were hesitant to include functional assessment at every visit as a key performance indicator in JIA. ³²

Third, our inability to detect trajectories with persisting moderate or severe functional impairments does not mean that no Canadian children with JIA experience substantial ongoing impairments. Despite modern treatments, many pediatric rheumatologists can recall at least one patient who developed significant disability despite best efforts. The number of such patients may be too small for LCTA to detect their trajectories, or they may come from marginalized communities with limited participation in the CAPRI Registry. Additionally, because most patients were followed for less than four years and the posterior probabilities of assignment to a trajectory were often below 0.7, caution should be exercised in interpreting LCTA results.

In conclusion, this longitudinal assessment of trajectories of physical function in children with JIA in Canada diagnosed in 2017 to 2024 show that shortly after diagnosis, patients have different levels of functional impairment related to their JIA category and they improve in function during the first year and subsequently stabilize. We identified five JIA data‐driven physical function trajectories with about one in three children having mild persisting impairments and no trajectories with persisting moderate or severe impairments. At baseline, multiple characteristics differed across the five trajectories, with patients with more involved joints and more severe symptoms having less favorable trajectories. If the KDS is available at baseline, a KDS score ≥1.5 would identify both children with current significant impairments and about 79% of children who subsequently develop persisting mild functional impairments. This KDS cutoff may help identify patients who may benefit from early, intensive targeted rehabilitation intervention.

AUTHOR CONTRIBUTIONS

All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Guzman confirms that all authors have provided the final approval of the version to be published and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements.

Supporting information

Disclosure Form:

Data S1. Supporting Information.