Pattern Recognition in Functional GI Disorders: 18 Distinct Symptom Clusters

Functional gastrointestinal disorders (FGIDs) affect approximately 40% of the global population, yet the Rome IV diagnostic criteria classify patients into a limited set of syndromes that poorly predict treatment response. This study applied unsupervised machine-learning cluster analysis (k-means with hierarchical validation and bootstrap resampling, n=1,000 iterations) to a multicentre patient registry of 5,814 adults with FGID diagnoses across 14 countries. Each patient was characterised using 67 symptom variables collected via validated standardised questionnaire covering gastrointestinal, psychological, and quality-of-life domains.

Analysis identified 18 statistically robust and clinically distinct symptom clusters (mean silhouette coefficient 0.71), substantially more granular than the 6 Rome IV categories. These 18 clusters explained 73% of treatment response variance across dietary, pharmacological, and psychological interventions — compared to 41% for Rome IV classification alone (p<0.001 for incremental variance explained). Notably, 6 clusters demonstrated highly specific and differential treatment responses that Rome IV groupings failed to distinguish: low-FODMAP restriction, tricyclic antidepressants, and pelvic floor physiotherapy each produced markedly superior outcomes in specific cluster subgroups. This 18-cluster taxonomy provides a superior framework for precision FGID management and forms the direct scientific basis for GutIQ's symptom pattern recognition engine.