Constipation Predominant Irritable Bowel Syndrome and Functional Constipation Are Not Discrete Disorders: A Machine...

Area of Science:

• Gastroenterology
• Medical Informatics
• Clinical Diagnostics

Background:

• Chronic constipation is clinically divided into irritable bowel syndrome with constipation (IBS-C) and functional constipation (FC).
• This classification assumes distinct pathophysiological differences and multiple clinical characteristics between IBS-C and FC.
• The validity of this syndromic segregation requires statistical and clinical validation.

Purpose of the Study:

• To test the assumption that IBS-C and FC are distinct syndromes using machine learning.
• To compare the diagnostic accuracy of models based on single versus multiple clinical features.
• To statistically formalize the grounds for syndromic segregation of IBS-C and FC.

Main Methods:

• Machine learning applied to a prospective cohort of 768 comprehensively phenotyped patients with chronic constipation.
• Development and comparison of 'unisymptomatic' (abdominal pain) and 'syndromic' (multiple features) diagnostic models.
• Statistical clustering of clinical characteristics to identify diagnostic structures.

Main Results:

• Unisymptomatic models using abdominal pain achieved near-perfect discrimination between IBS-C and FC (AUC 0.97).
• Syndromic models did not significantly improve diagnostic accuracy (P > 0.15).
• Models excluding abdominal pain performed at chance level (AUC 0.56), and no diagnostic structure was found via clustering.

Conclusions:

• IBS-C and FC may represent a single syndrome varying primarily by the presence of abdominal pain, rather than distinct entities.
• Current diagnostic criteria may create a self-fulfilling distinction based on abdominal pain.
• Findings impact clinical trial recruitment, disease classification, and management guidelines for chronic constipation.