Identification and validation of functional gastrointestinal disorder subtypes using latent class analysis: a population-based study

Objective: Attempts to categorize distinct functional gastrointestinal disorders based on reported symptoms continue but symptoms frequently overlap. The study objective was to use latent class analysis (LCA) which accommodates both continuous and discrete manifest variables to determine mutually exclusive subgroup assignments of a population-based sample using gastrointestinal symptom and patient data. Materials and methods: A validated bowel disease questionnaire and somatic symptom questionnaire were mailed to an age and gender stratified randomly selected community sample. Responses to the symptom questions were dichotomized as frequent vs. infrequent based on Rome IV criteria. A LCA model was developed using a calibration subset and the results applied to the validation subset. Results: There were 3831 total respondents (48%) with 3425 having complete data. The LCA algorithm was run for each of 10 (random) splits of the dataset and 2–6 latent classes were specified. Using the values of Akaike’s Information Criterion coefficient c to determine fit of the data, 4 latent classes yielded better values resulting in four subgroups: ‘asymptomatic,’ ‘upper’ abdominal symptoms, ‘lower’ abdominal symptoms, and ‘mixed’ (upper and lower abdomen). Conclusions: Latent class analysis identified 4 groups based on symptoms. This approach resulted in differentiation by anatomical region rather than the Rome IV classification of symptoms.