Characterization of active joint count trajectories in juvenile idiopathic arthritis

Abstract

Background/Purpose: To describe the patterns of longitudinal disease activity (active joint count (tender and swollen joints) in juvenile idiopathic arthritis (JIA) and to examine the association of clinical and laboratory characteristics with these patterns. Methods: A retrospective cohort study at two Canadian centres was performed. The longitudinal patterns of active joint counts were described using latent curve growth analysis. This method is ideally suited to a population whereby the underlying hypothesis is that the population is comprised of (unobserved) subpopulations. Latent curve growth analysis aims to classify individuals into statistically distinct groups based on individual response patterns so that individuals within a group are more similar than individuals between groups. The trajectory classes are each defined by a longitudinal growth curve. The association of baseline characteristics with class membership was performed by conducting a test of mean difference across classes for continuous variables and by comparing proportions for categorical variables. Results: Data were analyzed on 659 children diagnosed with JIA between 1990/03-2009/09. The median age at diagnosis was 10.00 (IQR 3.67-13.39), 61% (402/659) were female and 45% (286/629) were ANA positive. The distribution of the ILAR diagnoses were as follows: systemic (7%), oligoarthritis (36%), polyarthritis (RF negative) (13%), polyarthritis (RF positive) (4%), psoriatic arthritis (8%), enthesitis-related arthritis (20%) and undifferentiated (12%). A maximum of 10 years of follow-up data was included in the longitudinal analysis. The 659 patients were classified into 5 statistically different patterns of longitudinal active joint count (AJC) profiles using a latent curve growth analysis. 44% of patients were in group 1 characterized by a low initial AJC (mean 0.9) following by a decrease in joint count, 18% in group 2 - minimal to no active joint disease throughout course (mean 0.3), 19% in group 3 - moderate persistent AJC (mean 3.8), 10% in group 4 - initial mean AJC 4.9 followed by an increase in AJC at 5 years (mean 9.7) and finally 10% in group 5 characterized by an initial polyarthritis (mean 14) followed by a decline in AJC. The baseline characteristics of participants stratified by trajectory were statistically significantly different for all variables considered (age, sex, diagnostic delay, ANA positive, HLA-B27 positive, systemic fever, lumbosacral back pain, family history of HLA-B27 associated disease). Conclusion: This study was a successful application of a novel approach to longitudinal growth curve modeling to identify distinct trajectories of disease activity in JIA. The trajectories identified were statistically and clinically distinct from the ILAR subtypes. Identification of patterns of disease course is important in working towards the development of an outcome-based classification system in JIA.