Learning Longitudinal Patterns and Subtypes of Pediatric Crohn Disease Treated With Infliximab via Trajectory Cluster Analysis

Background: The objective of this study is to identify subgroups of pediatric Crohn disease (CD) who had differential responses to the infliximab treatment through trajectory cluster analysis of disease activity using data from electronic health records.

Methods: We conducted a retrospective study of 295 pediatric patients with CD who had been treated with infliximab for a minimum of one year at the Center for Inflammatory Bowel Disease at The Children's Hospital of Philadelphia between January 2010 and December 2017. The evolution of disease was described, and subgroups of patients were identified using trajectory analysis of longitudinal data of C-reactive protein (CRP). We compared patient characteristics, biomarker for disease activity, and long-term surgical outcomes across subgroups. Cox regression models were used to evaluate the added value of the subgroup classification to baseline phenotype and location in prediction of long-term surgical outcomes.

Results: We identified three subgroups of patients with differential relapse-and-remission profiles (n = 33, 65 and 197 from subgroup 1 to 3), which represented patients with a higher risk of infliximab non-response, with infliximab response but with occasional disease flares, and patients with long-term response. Patients with the best treatment response had a significantly lower frequency of complicated disease phenotypes (P = 0.01), including perianal involvement (P = 0.05), lower baseline CRP (P < 0.01) and calprotectin (P = 0.01), and lowest risk of IBD-related gastrointestinal surgery within 10 years of starting treatment (P < 0.01).

Conclusions: Readily available longitudinal data from electronic health records can be leveraged to provide deeper characterization of treatment response in pediatric CD.