Introduction

Diseases of the gastrointestinal tract affect 60-70 million people in the United States [1], and the total expenditure in 2015 for these illnesses was 135.9 billion dollars - greater than for other common diseases and likely to continue increasing. Abdominal pain alone accounted for $10.2 billion of the total costs [2]. GI afflictions and altered bowel habits affect the quality of life, social functioning and can result in considerable loss of productivity [3]. Early identification of appropriate markers for functional bowel disorders and a personalized therapeutic approach may help to reduce healthcare costs and improve quality of life.

Functional gastrointestinal disorders are conditions that present as normal upon examination of the GI system but still result in poor GI motility - primarily with symptoms in the middle or lower gastrointestinal tract [4]. These disorders include Irritable Bowel Syndrome (IBS), bloating, constipation, diarrhea, gas, and dyspepsia [4,5], among others. Although the pathophysiology of FGIDs is often complex due to their multifactorial nature, they are frequently encountered in both primary care and gastroenterology settings. FGIDs are thought to encompass bidirectional dysregulation of the gut-brain interaction via the gut-brain axis, gut microbial dysbiosis, visceral hypersensitivity, altered mucosal immune function, and abnormal gastrointestinal tract motility [6,7], and women appear to be afflicted at least twice as frequently as men [8].

Irritable Bowel Syndrome (IBS) is the most common functional bowel disorder and is prevalent in the range of 5–25% of the population and accounts for 36% of visits to a gastroenterologist. It is a gastrointestinal tract disease characterized by abdominal pain, altered bowel habits, and affected quality of life, but the absence of demonstrable organic disease is not considered life-threatening. IBS affects 10-15% of the US adult population [2,8]. Despite being a non-life-threatening disease, IBS can cause considerable pain and discomfort, and treatment often requires lifestyle changes and medication. The most common symptoms of IBS are chronic diarrhea - IBS-D (⅓ of IBS patients), chronic constipation - IBS-C, or both - IBS-M. Patients with IBS have a lower reported quality of life than sufferers of GERD and asthma [8].

Frequent and recurrent pain in the abdominal region not necessarily attributable to gut function is often referred to as Functional Abdominal Pain [14]. Children with functional abdominal pain were found to be significantly more likely to be obese [15]. FAP symptoms are not as common as other FGIDs and are not necessarily associated with food intake or passage but rather with psychiatric disorders. Management can involve psychotherapy and pharmaceutical interventions encompassing psychiatric regimens like anti-depressants, anticonvulsants, and treatments for other psychiatric disorders [16]. Functional bloating, on the other hand, is typically unlinked to psychiatric or organic causes [9]. Instead, it is a recurrent sensation of abdominal distention that may or may not be associated with measurable distention. Most people who report bloating have functional gastrointestinal disorders, and up to 96% of IBS patients report bloating. Bloating is 2x more commonly reported in women than men and typically worsens after meals [10,11]. Malabsorption of simple and complex carbohydrates and dietary fiber are commonly associated with both gas and bloating [6]. Gas is the byproduct of the bacterial fermentation of carbohydrates and protein in the large intestine, resulting in changes to the gut microbiome, increased short-chain fatty acids, and increased gas, diarrhea, abdominal pain, and bloating [9]. The exact connection between gas and bloating is not fully understood, but although intestinal gas may contribute to bloating, bloating does not necessarily result from more gas. Management often involves dietary modification, gut microbiome modulation, and lifestyle alteration [9].

Functional constipation is a functional bowel disorder that does not meet IBS criteria but presents as persistently difficult, infrequent, or seemingly incomplete defecation [10]. Chronic constipation occurs in up to 27% of people and affects all ages [10]. In women, constipation and associated straining to evacuate causes vaginal prolapse. Women undergoing prolapse surgery should understand the importance of correcting and managing constipation to decrease the risk of recurrence of the vaginal prolapse [12]. Depending on specific tests, constipation may be defined as (1) straining, hard stools, unproductive movements, infrequent stools, or incomplete evacuation; (2) less than three bowel movements per week OR daily stool weight less than 35 g/day OR straining for more than 25% of the period; (3) lengthy whole-gut or colonic transit. Surprisingly, stool frequency appears to have just a slender relationship with colonic transit and there are usually no demonstrable physiological abnormalities [10].

Functional diarrhea (FD) is a functional gastrointestinal disorder characterized by chronic or recurrent diarrhea not explained by structural or biochemical abnormalities of the gut. Functional diarrhea is characterized as the passage of loose or watery stools without abdominal pain or discomfort [10]. Unspecified diarrhea was reported by 4.8% of people throughout the US, but its duration and frequency are uncertain. Diarrhea is one of the most commonly reported symptoms for consulting a gastroenterologist and is also a common presenting issue among many patients in general practice [10]. Treatment of FD depends on establishing a correct diagnosis and needs to be distinguished from diarrhea-predominant irritable bowel syndrome (IBS-D) and other organic causes of chronic diarrhea. Once a physician has established the diagnosis, aggravating factors will need to be identified and eliminated, possibly including physiologic factors (e.g., small bowel bacterial overgrowth), psychological factors (e.g., stress and anxiety), and dietary factors (e.g., carbohydrate malabsorption) [13].

Diet is also considered an important trigger of gut-related symptoms. Poor nutrition, for example, consumption of highly processed and “fast” foods, has been implicated in FGID etiology [17,18]. Conversely, a Mediterranean diet has been associated with a lower prevalence of FGID [19]. A dietary therapeutics approach, for example a low-FODMAP diet in which rapidly fermentable carbohydrates that are poorly absorbed by the gut are eliminated or avoided as much as possible, is a typical dietary protocol for IBS patients [20]. Low-FODMAP diets have been associated with relieving other FGID symptoms, although functional dyspepsia seems least responsive to such a regimen [21,22].

Dietary fiber, which increases mucosal protein production, is widely used to treat chronic constipation [23]. This fiber is then digested in the colon, providing a substrate for microbial fermentation and resulting in byproducts such SCFAs, which have pro-motility effects and help with stool bulk and gas transit in the colon [24]. Also related to diet and FGID is the concept of non-celiac gluten sensitivity. Despite lacking serological and histological markers for celiac disease, a subset of FGID patients report significant alleviation of symptomatology upon elimination of dietary gluten and re-experience these symptoms upon reintroducing gluten [26].

Hormonal events, and genetic, environmental, and psychosocial factors have all been implicated to varying degrees in FGID etiology [10,20]. Sleep disturbance, dysfunctional coping, and psychological disturbance can correlate with symptom exacerbation. However, the nature of the link between psychosocial factors and FGIDs is poorly understood [10].

Both gut biome and genetics are likely contributors to FGID etiology. Evidence indicates a vital genetic component to FGIDs as demonstrated by prevalence within families and more substantial concordance between monozygotic versus dizygotic twins [27,28]. Functional dyspepsia has been associated with specific genotypes, including those associated with the ‘G-protein beta 3 subunit 825 CC genotype’, those associated with smooth muscles of the GI tract, and others associated with abdominal pain [29,30]. One evidence of a mutation linked to IBS was replicated across two studies, and although the mutation is found in only approximately 2% of IBS patients, this finding indicates the influence genetics may have on IBS symptoms [31,32]. Functional constipation has also been associated with specific genes [30]. Much of the research around genome-wide association studies and FGIDs is hampered by small sample sizes; however, the study of the genetics of FGID is rapidly evolving. The gut biome is also extensively implicated in FGID and particularly IBS pathogenesis [34,35]. Both IBS and functional dyspepsia have been shown to arise in susceptible individuals following a course of acute onset gastroenteritis [36]. Recent studies have revealed dysbiosis of the gut microbiota in constipated patients compared with healthy controls, associated with suppressed intestinal motility by metabolites produced by intestinal bacteria [24,37]. Other research has elucidated several mechanisms playing important roles in IBS. A dysregulated gut-brain axis has been adopted as a suitable model for IBS, and poor gut microbiome diversity may contribute to the onset and exacerbation of IBS symptoms. Dietary fibre appears to influence the gut microbiota, encouraging the growth of beneficial probiotics while preventing pathogenic and obesogenic bacteria from overgrowing [24,25]. Although clinical trials, which have attempted to characterise the gut microbiota in IBS, do not yet allow for a causal role to be inferred, they do confirm alterations in both community stability and diversity [38]. Although large-scale genomic analyses of IBS patients are lacking [20], numerous candidate genes have been potentially implicated in IBS. However, most studies involve small sample sets, have not been independently replicated, or are otherwise not robust. What seems clear so far is that most IBS sufferers either share several common gene variants that each nominally contribute toward the overall risk of the disease, or, for a subset of sufferers, a few highly penetrant alleles are likely the significant risk factors. Given that IBS spans both complex polygenic conditions and rare single-gene forms, it evidences the need for different strategies to identify these genetic factors.

Our aim for this study was to assess how effective a digital therapeutics intervention personalized on genomic SNPs and gut microbiome signals were at reducing symptomatology of FGIDs. Additionally, we aimed at building statistical models to describe the impact of each predictor (demographics, genomic SNPs, and gut microbiome) on the likelihood of a subject presenting or not FGIDs, and then model reduction of summative symptom severity (IBS, diarrhea, constipation, bloating, gas, and abdominal pain) as well as reduction of symptom severity in IBS, diarrhea, and constipation.

Material and Methods

Subject enrollment

Subjects were recruited from those who achieved 5% or more body weight loss when enrolled in the Digbi Health personalized digital care program, which delivers both human and digital personalized coaching for lifestyle changes, including diet and exercise, informed by demographics, genome SNP, and gut microbiome data (See supplementary methods for details). Only those subjects who had retrospectively responded to questions about symptomatology of their FGIDs and other comorbidities at the start of the program and after successful weight loss were included in the study (Figure 1). The average number of days for participants in the program at the time of the survey was 84.26 days. The final cohort studied in this manuscript included 177 subjects with either genome SNP (n=169) or baseline gut biome (n=168) or both analyses performed. Subjects were divided into two groups for comparisons and models: those who reported any of 6 FGIDs (IBS, diarrhea, constipation, bloating, gassiness, and cramping) at baseline or at the time of survey (n = 104), and those who reported no FGIDs (n=73).

• Download figure
• Open in new tab

Figure 1.

Study design flowchart.

Results

3. Subjects reported a reduction in severity of FGID related symptoms over the course of treatment

The proportion of subjects who experienced at least one point improvement in symptom severity ranged from 75.32% for diarrhea to 90.63% for bloating (Figure 2). Improvement in summative severity across the 6 FGIDs was seen by 89.42% of respondents, with an average summative reduction of 51.17% (Wilcoxon signed-rank test, p=4.89e-17*). The improvement in FGID symptomatology over the course of digital therapeutics intervention (percent summative reduction) was not correlated with percent weight loss (Kendall = 0.12, p=0.91), age (Kendall = -0.66, p=0.51), or gender (Wilcoxon sum rank, p=0.809).

• Download figure
• Open in new tab

Figure 2.

Self-reported symptom severity change (number of levels increased/decreased severity) in functional gastrointestinal disorders from baseline to time of survey.

Individually, we noted an average 45.93% reduction in the severity of IBS (Wilcoxon signed-rank, p=1.61e-08*), 61.01% average reduction in the severity of bloating (Wilcoxon signed-rank test, p=2.97e-16*), 38.55% average reduction in the severity of gassiness (Wilcoxon signed-rank, p=6.33e-12*), 61.69% average reduction in the severity of cramping/belly pain (Wilcoxon signed-rank, p=4.33e-12*), 37.54% average reduction in the severity of constipation (Wilcoxon signed-rank, p=3e-09*) and a 48.97% average reduction in the severity of diarrhea (Wilcoxon signed-rank, p=1.65e-07*).

Discussion

Out of 177 subjects enrolled in this study who successfully lost 5% or more body weight through a digital therapeutics program, 104 presented one or more FGIDs. These FGID sufferers were significantly different from the non-FGID group in terms of gender and BMI at the time of sampling (Table 1). Additionally, gender was significantly associated with the composition of baseline gut microbiome samples in these subjects (Table S2). Thus, using gender as a demographic variable, we trained logistic regression models to differentiate FGID status using D+G, D+M, and D+G+M variables. The models confirmed that the female gender was associated with a higher prevalence of FGID as seen in the logistic regression D+G model, where females were 3.26 times more likely to be FGID sufferers than males while holding genomic predictors constant (Table 2). This is in accordance with what has been reported elsewhere [54]. Additionally, when baseline microbiome was added to the model (D+G+M model), females were on average 2.26 times more likely to suffer from FGID than males while holding constant both genomic and microbial predictors (Table 4). Thus, some of the gender association in the D+G model is explained by microbiome variables in the D+G+M model, reinforcing the role of gender in shaping the baseline gut microbiome of subjects [55].

Of note, the SNP: gluten sensitivity (rs2187668, risk allele T) was seen to be strongly associated with FGID status in our cohort. This SNP variant is identified as HLA-DQ2.5 and has been reported as one of the most common HLA-DQ2 haplotypes associated with celiac disease [56]. Interestingly, the D+G+M model did not select any genomic variables and was identical to the D+M model (same variables but slightly different odd ratios), and not surprisingly, the pseudo R² scores for these models are similar (0.227 for D+G+M vs. 0.220 for D+M). These models have pseudo R² scores higher than for the D+G model (0.089), indicating that baseline microbiome better-classified participants having FGID than models based on genomic predictors and that the addition of SNPs did not improve classification of FGID by gender plus baseline microbiome. Many of the microbiome taxa variables identified in the models have already been reported in the literature associated with FGIDs. Previous studies show a strong association of the Ruminococcus torques group with FGIDs [57].

We then investigated the change in reported symptom severity over the course of the digital therapeutics program. In total, 89.42% of subjects experienced improvement of severity of at least one FGID. This improvement in symptom severity was not correlated with percent weight loss, gender, or age. Reduction in symptom severity was significant for all six FGIDs investigated individually and for summative symptom severity (all of them together). When we modeled the reduction in summative FGID symptom severity over the course of digital therapeutics intervention, we identified the D+G+M linear model as the best fitting, with an adjusted R² of 0.442, compared with 0.124 for the D+G model and 0.318 for the D+M model. Two genomic SNPs and eight microbial taxa were the significant predictors in the best model for these participants. Thus, in our cohort, gender and baseline microbiome best classified subjects with or without FGIDs, whereas a combination of genomic SNPs and microbiome variables (but not gender) best-modeled reduction in summative FGIDs symptom severity.

We then looked at the reduction of symptom severity for three functional bowel disorders of our interest: IBS, constipation, and diarrhea. Interestingly, for IBS, the best fit model was the one containing D+G+M variables, whereas, for constipation and diarrhea, the best fit models were those only containing D+M variables. When analyzing the variables found significant to the best fit models for this cohort, we identified several genomic SNPs and microbial taxa that are shared across two or more models. SNP rs7775228 (risk allele C) was associated in the linear D+G+M model with an increase in summative symptom severity (Table 5), as well as in the linear D+G+M model with an increase in IBS symptom severity (Table 6). Interestingly, despite not being the best fit model, it also was associated in the linear D+G model with an increase in diarrhea symptom severity (Table S9) and associated in the linear D+G+M model with reduction of constipation symptom severity (Table 7). In addition to its association with gluten sensitivity [58], which is why we include this SNP as part of our care protocols, rs7775228 is involved in seasonal allergic rhinitis and as a protein biomarker for inflammation [58,59].

In terms of the microbial taxa shared across two or more models, genus Fusicatenibacter was associated with a greater prevalence of FGID in the D+G+M logistic model of this cohort (Table 4) and associated with an increase in IBS symptom severity in the D+G+M linear model (Table 6). Genus Intestinimonas was associated with a reduction in summative FGID symptom severity in the D+G+M model (Table 5) and a reduction in diarrhea symptom severity in the D+M model (Table S10). Genus Megasphaera was noted to be strongly associated with a reduction in summative FGID symptom severity in the D+G+M linear model of Table 5 and a reduction in IBS symptom severity in the D+G+M linear model of Table 6. Interestingly, genus Lactobacillus appears associated with a reduction in diarrhea symptom severity in the linear D+M model of Table S10, an effect that has been amply demonstrated in the literature [60], supporting the validity of our methods. Moreover, these bacteria, specifically, Fusicatenibacter, Megasphaera, and Intestinimonas, were either negatively associated with FGID status or were associated with a reduction in the severity of FGID symptoms are previously reported to be Short Chain Fatty Acids (SCFAs) producers [61–64]. There is ample evidence of the role of SCFAs in improving gut integrity, which plays an essential role in maintaining mucosal homeostasis [65].

Our analysis also revealed some bacterial signatures that were associated with FGID status. Desulfovibrio was observed to possess a significant association with an increase in FGID symptomatology (Table 5). Previous reports suggest that bacteria belonging to the genus Desulfovibrio generate H₂S gas via a dissimilatory sulfate reduction pathway, leading to inflammatory gut disorders [66,67]. We also noted the association of Akkermansia with FGID status (Table 1). Although this bacterium is suggested to have beneficial associations with the gut health, a few studies have reported its inverse correlation with reduction of abdominal pain [68].

Ruminococcus torques group appears associated with FGID status in the D+G+M logistic model of Table 4 and associated with a reduction in summative FGID symptom severity in the D+G+M model of Table 5. Despite not being the best fit model, this taxon also was associated with a reduction of diarrhea symptom severity in the linear D+G+M model of Table 8. Different Ruminococcus torques subgroups have been associated in the literature with IBS-D, IBS-M, and Crohn’s disease subjects [57]. Additionally, genus Terrisporobacter, associated with inflammation and gut dysbiosis [69], was associated with FGID status in the D+G+M logistic model (Table 4). Collectively, these findings indicate a potential of gut microbial profiling not only for predicting current gastrointestinal health but also education in FGID related symptoms.

This study has some limitations that are important to note. First, the descriptive modeling exercise performed in this work is the best fit for the cohort analyzed here and is not intended to infer for the larger population. In particular, we aimed to investigate which demographic, genomic, and baseline microbiome predictors improved the fit of the models, along with the magnitude and direction of their association with FGID. Second, whereas we used all microbiome taxa present in the baseline gut microbiome samples (n=105 after the filters imposed), for genomic SNPs, we selected only markers that are used to inform diet and lifestyle interventions of subjects under the Digbi Health program, specifically those associated with intolerances and allergies (n=14). So, the fact that microbiome markers almost always outperformed genomic SNP markers may be due to the markedly different dataset sizes. Third, inclusion criteria in this study did not consider factors known to influence the microbiome composition (probiotic or antibiotic usage) or other comorbidities (musculoskeletal pain, skin conditions, hypothyroidism, diabetes, cholesterol, hypertension, and mental health) that may confound the results presented. And fourth, the survey instrument utilized was an ad-hoc questionnaire that asked participants to rate their symptom severity for different FGIDs on a scale of 1 to 5 and was not a validated clinical instrument. The survey was performed retrospectively for both time points after successfully achieving 5% or more body weight loss.

Despite the above limitations, the digital therapeutics care provided to subjects, informed by genetic and baseline gut microbiome and their interaction with participant’s lifestyle, seemed to effectively reduce symptom severity of FGIDs, including IBS, diarrhea, and constipation. Our earlier study supported the use of this care as a therapy for insulin resistance [70], empowering subjects to manage their inflammation by awareness of the impact of processed foods and foods to which they are sensitive as per their genomic SNPs and microbiome results. Dietary fiber coaching also results in increased vegetable diversity and quantity. Whereas further research is required to better understand the effect of different components of the care (e.g., fiber types) on modulating the microbial taxa and genomic SNPs identified in the models and their corresponding effect on reduction of FGIDs symptom severity, this preliminary retrospective study generates testable hypotheses for associations of a number of biomarkers in the prognosis of FGIDs symptomatology. Moreover, the methods presented add to the existing set of tools (e.g. [71]) that can be readily implemented to understand the role that genetics and gut microbiome play on disease etiology.

Authors’ statements

Supplementary Methods

Supplementary tables