In conclusion, using a Gaussian mixture model analysis, we identified six different IBS subgroups, dominated by the presence or absence of diarrhea and/or constipation and non-GI somatic and psychological comorbidity, in line with previous studies. We also demonstrated the importance of these subgroups based on differences in patient-reported outcomes as well as gut physiology, in particular oro-anal transit time. These findings highlight the importance of a biopsychosocial approach in patients with IBS.
For this study, we combined data from six different study cohorts in which similar data was collected in the period between 2010 and 2021 (Dnr 731/09, Dnr 988/14, NCT03869359, NCT02970591, NCT05182593, and NCT02107625). Individuals were recruited using different strategies across the studies, ranging from referral from primary and secondary care, self-referral, or advertisements in newspapers, to university buildings and social media. Data from one of the studies was previously used for similar subgroup analyses as in the current study. The six studies were performed at our specialized unit for neurogastroenterology at Sahlgrenska University Hospital, Gothenburg, Sweden. IBS diagnosis was confirmed by MS, HT, or a specially trained gastroenterology resident, according to the Rome criteria at time of inclusion (Rome III, 2010-2016; Rome IV, 2016-2022). IBS subtypes were based on stool diaries and reported according to the Rome III criteria (IBS-D, IBS-C, IBS-M and IBS-U). Although the exclusion criteria varied slightly across studies, they generally excluded participants with (severe) comorbidities, particularly GI diseases that could explain GI symptoms. Patients with completely missing data on the Gastrointestinal Symptom Rating Scale IBS (GSRS-IBS) (used for the clustering) were not included in this study.
The GSRS-IBS is a 13-item questionnaire used to assess how bothersome symptoms were over the preceding week (all GSRS-IBS items available in n = 757). All 13 questions were used as individual items with a seven-point Likert scale, from 1 (no discomfort at all) up to 7 (very severe discomfort). These questions included "abdominal pain", "pain/discomfort relieved by defecation", "bloating", "passing gas", "visible abdominal swelling (abdominal distention)", "constipation" (referring to infrequent bowel movements), "hard stools", "diarrhea" (referring to frequent bowel movements), "loose stools", "urgency", "incomplete bowel emptying", "early satiety", and "fullness". The average of the 13 items was reported as the GSRS-IBS total score.
The IBS severity scoring system (IBS-SSS) is a questionnaire to assess IBS symptom severity in patients with IBS with the score calculated based on 5 questions (n = 739). Scores range from 0 (no symptoms) to 500 (maximum severity). The IBS symptom severity can be divided into subgroups of mild (75-174), moderate (175-299), or severe (≥300) IBS symptoms.
Depending on the cohort, a 7-, 10- or 14-day diary was used to calculate the average stool consistency and frequency. For stool consistency, the Bristol Stool Form Scale was used, ranging from 1 (separate hard lumps) up to 7 (watery, no solid pieces) (n = 730). Stool frequency was reported as the average number of stools per day.
The Patient Health Questionnaire (PHQ) 15 is a 15-item questionnaire used to assess the severity of somatic symptoms. After excluding the 3 GI-related questions, the remaining 12-item version is referred to as the PHQ-12. We used all individual questions of the PHQ-12, except the question regarding menstrual cramps or other problems with the period since this is only applicable to women of reproductive age (total score available in n = 724). All questions were scored on a three-point Likert scale, from 0 (not bothered at all) up to 2 (bothered a lot) and included "back pain", "pain in arms, legs, or joints (knees, hips, etc.)", "headaches", "chest pain", "dizziness", "fainting spells", "feeling the heart pound or race", "shortness of breath", "pain or problems during sexual intercourse", "feeling tired or having low energy", "trouble sleeping". In one cohort (n = 29), the response to question "trouble sleeping" was not available. To address this, the response to the question regarding trouble with sleeping, included in the Central Sensitization Inventory, was modified and used as a substitute.
The Hospital Anxiety and Depression scale (HADS) is a 14-item questionnaire used to assess psychological distress in non-psychiatric patients (n = 736). All 14 questions were scored based on a four-point Likert scale (0-3). Subscale scores for symptoms of anxiety and depression were calculated (both ranging from 0 to 21), with higher scores indicating severe symptoms.
The Visceral Sensitivity Index (VSI) is a 15-item questionnaire used to assess GI symptom-specific anxiety, developed for and validated in patients with IBS (n = 743). All 15 questions use a six-point Likert scale (1-6). The total score was used with higher scores indicating more severe GI-specific anxiety.
The Multidimensional Fatigue Inventory (MFI) is a 20-item questionnaire used to assess levels of fatigue (n = 694). All 20 questions use a five-point Likert scale (1-5). The MFI consists of 5 different dimensions: general fatigue, physical fatigue, reduced activity, reduced motivation, and mental fatigue. Higher scores indicate increased severity of fatigue.
The IBS quality of life questionnaire (IBSQOL) is a 30-item questionnaire used to assess IBS-specific quality of life (n = 420 of whom 39 with some missing data). All 30 items use a five- or six-point Likert scale. The IBSQOL consists of 9 different dimensions: emotional functioning (n = 420), mental health (n = 420), sleep (n = 420), energy (n = 420), physical functioning (n = 419), food (n = 420), social role, physical role (n = 420), and sexual relations (n = 381). The score is transformed into a 0-100 scale for each dimension, with higher scores indicating a better quality of life. The average of the 9 dimensions was calculated for each patient and used as overall IBS-specific quality of life.
The oro-anal transit time was assessed using our previously validated protocol using radiopaque markers (n = 301). In this protocol, the subjects ingested 10 radiopaque markers every morning for five consecutive days. On day 6, they ingested five radiopaque markers at 8 am and another five at 8 pm, to be able to investigate those patients with a rapid GI transit. On the morning of day 7, under fluoroscopic imaging, the number of radiopaque markers were counted and divided by ten to calculate the oro-anal transit time with days as unit.
Rectal sensitivity data was available in 2 cohorts, with different rectal barostat protocols used (any rectal barostat data available, n = 283). In the first method, a polyethylene balloon was attached to a polyvinyl tube (Sherwood Medical). The balloon was inserted into the rectum and connected to a computer driven electronic barostat (Dual Drive Barosta, Distender Series II; G&J Electronics Inc.) that controlled the inflation. First, a habituation sequence was performed with a 4 mmHg increase every 15 s until 20 mmHg or the discomfort threshold was reached. Thereafter, the protocol started from 0 mmHg with a 4 mmHg increasement every minute until the subject reported pain or a maximum pressure of 60 mmHg was reached. The thresholds (mmHg) for first sensation, urge to defecate, discomfort, and abdominal pain were reported.
In the second method, a polyethylene balloon was inserted into the rectum and connected to the rapid barostat bag pump (Mui Scientific) that controlled the inflation. This is a battery-operated air compressor, which performs a ramp distension of the bag with 120 ml/min. The balloon was first inflated up to 40 mmHg intrabag pressure, with the corresponding volume reported as rectal capacity. Thereafter the bag was deflated and then reinflated with 120 ml/min, until the thresholds of first sensation, urge to defecate, and discomfort were reached. Pain was not measured as separate threshold during this procedure. This procedure was repeated for a second time. The thresholds (mmHg) during this repetition were used in this study. To be able to combine the data, Z-scores were calculated for both methods and used in the final analysis.
During the lactulose-nutrient challenge test (LNCT), the participants ingested a combined lactulose (25 g) and nutrient (400 mL Fortimel Energy, 600 kcl, 49 E % carbohydrates, 35 E % fat, 16 E % protein; gluten-free and lactose free) drink after an overnight fast (n = 300). Our previous publications describe the LNCT in more detail. Hydrogen and methane production were measured in exhaled air pre- and postprandial, every 15 min for four hours (QuinTron Breath Tracker; QuinTron Instrument Company) as indirect measurements of fermentation by the gut microbiome.
Numbers, proportions, means with standard deviations (SD), and medians with interquartile ranges (IQR) were used for the descriptive analyses. In order to identify subgroups in the total sample, we deployed a Gaussian finite mixture model fitted by an expectation maximization algorithm as implemented in the Mclust package in R and as described extensively before. Briefly, we included the same variables as in our previous study to check the reproducibility of previous findings. These variables included the average stool frequency and consistency, all individual GSRS-IBS items, 11 PHQ-12 items, and the 2 HADS subscale scores (anxiety and depression). These variables were chosen based on previous data suggesting that symptoms of depression, anxiety and somatization (non-GI symptoms) are major contributors to a lower quality of life and increase bowel symptom burden in individuals with IBS and functional dyspepsia. The only difference from the previous model was that we excluded the PHQ-12 item on menstrual cramps or other problems with the period as a non-GI somatic symptom, since this is age- and sex-specific. First, missing data was imputed by R package MissMDA. After that, all variables were log-transformed, and prior to transformation, 1 was added to avoid undefined ln(0) values. Following log-transformation, the clustering was performed with the optimal number of clusters determined on the optimal Bayesian Information Criterion. Each individual was assigned to the cluster (subgroup) in which they had the highest probability of membership. All original scores included in the model were transformed to Z-scores for illustration purposes and depicted in radar plots, with zero representing the overall group mean. Based on the interpretation of the symptom pattern compared to the average of the total group, a name was assigned to each subgroup.
To test for differences in categorical variables between the newly formed subgroups, the chi-square test was used. Testing for differences of continuous variables was determined depending on if they were well-modeled by a normal distribution (Shapiro-Wilk test), and was followed up by using the Kruskal-Wallis test (post-hoc: Dunn's test with the Benjamini-Hochberg correction), or one-way ANOVA (no post-hoc tests performed in this study). For illustration purposes, all patient-reported outcomes and measurements of gut physiology were min-max rescaled to 0 as the lowest value and 1 as the maximal value, with the median rescaled value of each measurement depicted in Figs. 2 and 4.
To assess the production of breath hydrogen and methane production over time, we used linear mixed models. The models included the main effects of time of respectively hydrogen and methane (continuous variable), and subgroup-by-time interaction effects to assess differences in patterns over time between the six subgroups. Benjamini-Hochberg correction was applied to correct for multiple testing.
Analyses were performed in R version 4.4.0 and 4.3.3.1 and "Stats, version 4.3.2" and IBM SPSS version 29.0.0.0 software. A p-value of <0.05 was considered statistically significant.
Ethical approval was provided by the ethics committee of the University of Gothenburg, Sweden, and the studies were performed in accordance with the Declaration of Helsinki. All subjects gave verbal and written informed consent before participating.