Early Identification of Pediatric Inflammatory Bowel Disease Based on a Noninvasive Multivariable Predictive Model

Area of Science:

• Pediatric Gastroenterology
• Diagnostic Imaging
• Biomarkers

Background:

• Diagnostic delays in pediatric inflammatory bowel disease (IBD) negatively impact long-term outcomes.
• Noninvasive evaluation methods are crucial for early IBD detection in children.
• Current diagnostic approaches often involve delays, necessitating improved early identification strategies.

Purpose of the Study:

• To establish and validate an optimal model for noninvasive evaluation tests for early pediatric IBD identification.
• To develop a predictive model utilizing clinical symptoms, laboratory tests, and ultrasound findings.
• To reduce diagnostic delays and improve outcomes for children with IBD.

Main Methods:

• Retrospective development and prospective temporal validation design.
• Logistic regression model developed using data from 314 pediatric patients (103 IBD, 211 non-IBD).
• Incorporated IBD symptoms, routine labs, and transabdominal ultrasound (Limberg score >1, mesenteric fat, wall layering, lymph nodes).
• Validated in a prospective cohort of 66 children (19 IBD, 47 non-IBD).

Main Results:

• The ultrasound feature Limberg level >1 was the most valuable predictor.
• Key features included erythrocyte sedimentation rate, fecal calprotectin, C-reactive protein, hypoalbuminemia, and perianal abscess/fistula.
• The model achieved high accuracy: AUC 0.97 (internal) and 0.94 (external validation).
• Demonstrated good calibration in external validation (slope 0.86, Brier score 0.08).

Conclusions:

• A nomogram based on noninvasive factors can effectively identify children with IBD in early stages.
• Accessible noninvasive testing can facilitate early diagnosis of pediatric IBD.
• The validated model offers a promising tool for clinicians to improve pediatric IBD detection.