Disease activity classification based on routinely collected blood tests in a prospective paediatric inflammatory bowel disease inception cohort

We have shown that disease-activity clusters derived from routine blood tests in administrative data predicted disease outcomes in Crohn's disease (CD). We aimed to validate these findings on a prospective inception cohort. Hierarchical clustering grouped children newly diagnosed with CD by similarities in diagnostic blood tests and fecal calprotectin and ordered by median laboratory values. The clusters were then validated against constructs of disease activity (weighted pediatric Crohn's disease activity index, physician global assessment, and simple endoscopic score [SES]-CD), and by their ability to predict disease course. Among 147 children with CD, three clusters were formed (mild, moderate, and severe with progressing worsening of laboratory values). A stepwise increase was observed in disease-activity constructs (e.g., median SES-CD 8 [3-12], 14 [8-18], 16 [8-26]; p < 0.001). Higher cluster severity was associated with a complicated-disease-course (vs. mild: hazard ratio 2.68; 95% confidence interval 1.19-6.05). Clustering routinely collected blood tests can reflect disease activity in CD and predict disease course.