Data Mining Approach for the Characterization of Functional Bowel Disorders According to Symptom Intensity Provides a Small Number of Homogenous Groups

Abstract

Background/Aims: The aim of the present study is to evaluate if the intensity of the cardinal symptoms of functional bowel disorders could be used to identify homogenous groups of patients defined by the Rome criteria. Method: In this observational study, 1,729 consecutive outpatients (73% females) filled out the Rome III questionnaire and 10-point Likert scales for constipation, diarrhea, bloating (BL)/distension, abdominal pain (AP) during the week before the medical consultation. A Gaussian mixture model was used for clustering the patients according to the intensity of symptoms without a priori information, and a classification tree was constructed from this clustering. Data were analyzed using analysis of variance and logistic regression analysis. Results: According to the intensity of symptoms, the patients are divided into 8 groups named according to their main symptomatology: “painful constipation” (PFC), “mild pain constipation” (MPC), “painful diarrhea” (PFD), “mild pain diarrhea” (MPD), “mixed transit” (MT), “BL,” “AP,” and “nonspecific” (NS). The study of the relationship between the Rome III classification and this new grouping shows that irritable bowel syndrome (IBS)-constipation is associated with PFC, IBS-diarrhea with PFD and MPD, SII-mixed with MT, SII-unspecified with BL, functional constipation with PFC and MPC, functional diarrhea with MPD and NS, BL with “BL” and NS, nonspecific functional bowel disorders (FBD) with NS, and functional AP with “BL” and AP (p < 0.01 for all associations). Conclusion: A symptom intensity-based classification of FBD patients could simplify clinical phenotype, give homogeneous groups of patients, and could eventually be used by nongastroenterologists and in clinical research.