Analysis and Hierarchical Clustering of Blood Results before Diagnosis in Pediatric Inflammatory Bowel Disease

Abstract

Background: Pediatric inflammatory bowel disease (PIBD) is associated with a diagnostic delay. Blood tests are a routine part of the work-up in children with chronic abdominal symptoms (pain, diarrhea). Normal blood tests cannot exclude PIBD. We analyzed blood results at diagnosis over a 5-year period. Methods: Patients diagnosed from 2013 to 2017 were identified from the Southampton-PIBD database. Results were obtained up to 100 days before diagnostic endoscopy. Erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), albumin, hemoglobin, platelets, packed cell volume (PCV), white cell count (WCC), and alanine transferase (ALT) were analyzed. Hierarchical clustering was applied to normalized results. Results: Two hundred fifty-six patients were included (Crohn's disease [CD], 151; ulcerative colitis [UC], 95; IBD-unclassified, 10; median age, 13.48 years; 36.7% female). Hierarchical clustering of patients revealed novel groupings enriched for CD and UC, characterized by specific patterns of results. In PIBD, 9% presented with all normal blood tests, 21.9% with normal CRP and ESR. Abnormal results were seen in all tests (ESR, 56.4% of patients; CRP, 53.4%; albumin, 28%; hemoglobin, 61.9%; platelets, 55.6%; PCV, 64.6%; WCC, 22.7%; and ALT, 7.2%). Normal inflammatory markers were more common in UC compared with CD (UC, 34%; CD, 15.8%; P = 0.0035). UC (14.4% normal) presented with all normal results more frequently than CD (CD, 5.3%; P = 0.02). CRP, ESR, and platelets were significantly higher in CD compared with UC. Albumin and hemoglobin were significantly lower. Conclusions: Most cases of PIBD present with >1 abnormal blood result, although 1/11 patients presents with normal blood tests and 1/5 present with normal inflammatory markers. Hierarchical clustering offers the potential to produce novel groupings to inform disease categorization and best management.