A Low-FODMAP Diet Provides Benefits for Functional Gastrointestinal Symptoms but Not for Improving Stool Consistency and Mucosal Inflammation in IBD: A Systematic Review and Meta-Analysis

Background: A low fermentable oligosaccharides, disaccharides, monosaccharides, and polyols diet (LFD) is claimed to improve functional gastrointestinal symptoms (FGSs). However, the role of LFD in inflammatory bowel disease (IBD) patients with FGSs remains unclear. Objective: To systematically assess the efficacy of LFD in IBD patients with FGSs. Methods: Six databases were searched from inception to 1 January 2022. Data were synthesized as the relative risk of symptoms improvement and normal stool consistency, mean difference of Bristol Stool Form Scale (BSFS), Short IBD Questionnaire (SIBDQ), IBS Quality of Life (IBS-QoL), Harvey-Bradshaw index (HBi), Mayo score, and fecal calprotectin (FC). Risk of bias was assessed based on study types. A funnel plot and Egger’s test were used to analyze publication bias. Results: This review screened and included nine eligible studies, including four randomized controlled trials (RCTs) and five before–after studies, involving a total of 446 participants (351 patients with LFD vs. 95 controls). LFD alleviated overall FGSs (RR: 0.47, 95% CI: 0.33–0.66, p = 0.0000) and obtained higher SIBDQ scores (MD = 11.24, 95% CI 6.61 to 15.87, p = 0.0000) and lower HBi score of Crohn’s disease (MD = −1.09, 95% CI −1.77 to −0.42, p = 0.002). However, there were no statistically significant differences in normal stool consistency, BSFS, IBS-QoL, Mayo score of ulcerative colitis, and FC. No publication bias was found. Conclusions: LFD provides a benefit in FGSs and QoL but not for improving stool consistency and mucosal inflammation in IBD patients. Further well-designed RCTs are needed to develop the optimal LFD strategy for IBD.


Introduction
Inflammatory bowel disease (IBD), including ulcerative colitis (UC) and Crohn's disease (CD), is a group of chronic gastrointestinal diseases with frequent functional gastrointestinal symptoms (FGSs), such as abdominal pain and bloating. Prior preliminary cross-sectional research works reveal an overall prevalence of FGSs in 30% to 45% of IBD outpatients and a negative impact on both psychological wellbeing and quality of life (QoL) in the presence of these symptoms [1]. Higher rates of anxiety and depression and lower QoL scores were reported consistently in IBD patients with FGSs [1][2][3]. The etiology of these FGSs in IBD remains unclear, and the gastrointestinal damage or the psychological impact of IBD may be partially responsible in the process [4].
Trials, Chinese National Knowledge Infrastructure (CNKI), WanFang (Chinese) Database, with the last search update on 1 January 2022. Studies were screened without geographical and language restrictions. The search terms for retrieval in these databases were: 'FODMAP' OR 'FODMAPS' OR 'Fermentable, poorly absorbed, short-chain carbohydrates', OR 'Fermentable oligosaccharides, disaccharides, monosaccharides and polyols', AND 'inflammatory bowel disease' OR 'IBD' OR 'Crohn's disease' OR 'CD' OR 'ulcerative colitis' OR 'UC' or equivalent terms. Additionally, the retrieved references were screened manually to find the relevant potential literature.

Literature Screening
Eligible studies fulfilling the following criteria were included in our meta-analysis: (1) all relevant randomized controlled trials (RCTs) or before-after studies in the same patient; (2) a definitively established IBD diagnosis; (3) comparing LFD with a placebo diet or a usual diet (hereafter referred to as normal diet (ND)) or comparing pre-and postcontrast LFD; and (4) outcomes including overall and individual FGSs response, SIBDQ, IBS-QoL, GSRS, stool consistency, Mayo score for UC, HBi for CD, and FC. The exclusion criteria were presented as follows: (1) participants suffering from other digestive disorders; and (2) participants receiving multiple interventions simultaneously. Literature screening was carried out by two independent investigators (ZP and JY), and a third investigator (XL) resolved disagreements.

Data Extraction
From each included study, the following information was collected by two independent investigators (ZP and JY) using a standardized data extraction form: (1) general information: title, the first author, publication year, and the country of the study; (2) study information: study design, participants, intervention, duration of therapy, and outcome evaluation of FGSs; (3) baseline characteristics: total case/controls or cohort size, age range or mean age (standard deviation, SD), sex, and type of IBD; (4) outcomes: the number or percentage of patients with overall and individual FGSs improvement and normal stool consistency before and after the intervention; the mean difference (MD) of GSRS, IBS-SSS, SIBDQ, IBS-QoL, Bristol Stool Form Scale (BSFS), Mayo score for UC, HBi for CD, and level of FC. One investigator (ZP) was responsible for contacting the original author for complete data. All investigators participated in the discussion to resolve the dispute.

Assessment of Risk of Bias
The risk of bias assessment in the included RCTs was performed by two independent investigators using the Cochrane Risk of Bias Tool and Jaded scale with Review Manager (RevMan) (Version 5.3, Cochrane Collaboration, Oxford, UK) [31]. The investigators evaluated the quality of the non-RCTs included, according to the Methodological Index for Non-Randomized Studies (MINORS) [32]. The best possible score on this scale is 16 points.

Statistical Analysis
Data analysis was performed using RevMan 5 (Version 5.3, Cochrane Collaboration, Oxford, UK). Firstly, a chi-square test and the I 2 statistic were used to assess the heterogeneity of each study included. Among them, I 2 statistic means the percentage of total variability due to heterogeneity between studies. Secondly, according to the result of heterogeneity, the appropriate Mantel-Haenszel fixed-effects model or the DerSimonian and Laird random-effects model were selected to calculate the MD with 95% confidence intervals (CIs) for continuous data and the risk ratio (RR) with 95% CIs for dichotomous data. The random-effects model was used in case of high heterogeneity (p < 0.10 or I 2 statistic value > 50%); otherwise, the fixed-effects model was used (p > 0.10 or I 2 statistic value < 50%). In addition, a funnel plot and the Egger's test (using the Stata 15 software) were used to assess publication bias with a p value of <0.10 indicating statistical significance.

Characteristics of Eligible Studies
Initially, 476 articles were identified and screened via reviewing the title, abstract, and full text. Then, 467 articles were excluded for various reasons, such as non-human or non-original research or article with incomplete information, etc. Finally, nine studies were included for estimating the effect of LFD on IBD patients. A flowchart shown in Figure 1 presents the details of included studies and the selection process.
The random-effects model was used in case of high heterogeneity (p < 0.10 or I 2 statistic value > 50%); otherwise, the fixed-effects model was used (p > 0.10 or I 2 statistic value < 50%). In addition, a funnel plot and the Egger's test (using the Stata 15 software) were used to assess publication bias with a p value of <0.10 indicating statistical significance.

Characteristics of Eligible Studies
Initially, 476 articles were identified and screened via reviewing the title, abstract, and full text. Then, 467 articles were excluded for various reasons, such as non-human or non-original research or article with incomplete information, etc. Finally, nine studies were included for estimating the effect of LFD on IBD patients. A flowchart shown in Figure 1 presents the details of included studies and the selection process. Eventually, our meta-analysis was performed based on the inclusion of three prospective studies [20][21][22], one retrospective study [23], one study that included both prospective and retrospective components [24], and four RCTs [25][26][27][28], with a total of 446 IBD patients. Specifically, the five non-RCTs involved 256 IBD patients, and the four RCTs randomly divided IBD patients into the experimental group and the control group, involving 190 patients. Among them, IBD patients with ND, used as the control, contained 95 cases (nine patients participating in the cross-over trials were included in LDF group). Therefore, nine studies involving 351 LFD cases and 95 controls were analyzed in this Eventually, our meta-analysis was performed based on the inclusion of three prospective studies [20][21][22], one retrospective study [23], one study that included both prospective and retrospective components [24], and four RCTs [25][26][27][28], with a total of 446 IBD patients. Specifically, the five non-RCTs involved 256 IBD patients, and the four RCTs randomly divided IBD patients into the experimental group and the control group, involving 190 patients. Among them, IBD patients with ND, used as the control, contained 95 cases (nine patients participating in the cross-over trials were included in LDF group). Therefore, nine studies involving 351 LFD cases and 95 controls were analyzed in this meta-analysis. The baseline characteristics and data extraction from the included studies are outlined in Tables 1 and 2.    The visual analog scale score was used to measure overall gastrointestinal symptoms.

Overall Symptom Response
Considering the inconsistent definition standards for FGS improvement in different studies generally (Table 2), we analyzed the number of people suffering from FGSs before and after LFD intervention in non-RCTs or the LFD group and ND group in RCTs. As a whole, all the nine studies showed that LFD was associated with an improvement of

Overall Symptom Response
Considering the inconsistent definition standards for FGS improvement in different studies generally (Table 2), we analyzed the number of people suffering from FGSs before and after LFD intervention in non-RCTs or the LFD group and ND group in RCTs. As a whole, all the nine studies showed that LFD was associated with an improvement of

Degrees of Change in FGSs
Three studies assessed FGS changes using GSRS as the continuous variable, showing that LFD was associated with a reduction in total GSRS score (MD = −0.43, 95% CI −0.54 to −0.33, p = 0.000) (Figure 4a). Meanwhile, two studies assessed FGS changes using IBS-SSS,

Disease Activity
For UC, two studies reported the Mayo score, yet with no difference between the LFD group and ND group (MD = −0.32, 95% CI −1.09 to 0.45, p = 0.41) (Figure 5c). In contrast, three studies showed that LFD was associated with a reduction in HBi score for CD (MD = −1.09, 95% CI −1.77 to −0.42, p = 0.002) (Figure 5d).

Quality of the Included Studies
The overall risk of bias of four included RCTs was relatively low, as shown in Figure  6. Meanwhile, for the remaining five non-RCTs, according to the MINORS, four studies scored 14 points, and one study scored 12 points (Table 1).

Disease Activity
For UC, two studies reported the Mayo score, yet with no difference between the LFD group and ND group (MD = −0.32, 95% CI −1.09 to 0.45, p = 0.41) (Figure 5c). In contrast, three studies showed that LFD was associated with a reduction in HBi score for CD (MD = −1.09, 95% CI −1.77 to −0.42, p = 0.002) (Figure 5d).

Quality of the Included Studies
The overall risk of bias of four included RCTs was relatively low, as shown in Figure 6. Meanwhile, for the remaining five non-RCTs, according to the MINORS, four studies scored 14 points, and one study scored 12 points (Table 1). Nutrients 2022, 14, x FOR PEER REVIEW 10 of 15 Figure 6. Risk of bias summary of the four included randomized controlled trials (RCTs) [25][26][27][28]. (+, high-risk; ?, uncertain).

Publication Bias
No evidence of publication bias was found based on Egger's regression test, i.e., overall symptom response (p = 0.

Publication Bias
No evidence of publication bias was found based on Egger's regression test, i.e., overall symptom response (p = 0.

Discussion
This updated systematic review and meta-analysis included four RCTs and five beforeafter studies, with 446 participants in total. The study aimed to pool data from existing studies to examine whether LFD alleviates FGSs effectively in IBD patients. Additional data were extracted from existing studies and were used to uncover the efficacy of LFD on SIBDQ, IBS-QoL, stool consistency, Mayo for UC, HBi for CD, and FC in IBD patients. The present study found that LFD alleviated FGS, obtained higher SIBDQ scores, and reached remission or low disease activity in CD. However, there were no statistically significant differences in the efficacy of LFD on the IBS-QoL, stool consistency, Mayo for UC, and FC in IBD patients. It is worth noting that, in addition to assessing FGSs, the number of original studies and participants included is small. The credibility of these results remains unexamined, which requires large-scale clinical trials for further confirmation. Collectively, this meta-analysis suggested that IBD patients with FGSs may profit from LFD treatment with the assistance of a healthcare professional.
The primary outcome of this study was that LFD can improve FGSs in IBD. Symptom improvement was observed in bloating, wind or flatulence, borborygmi, abdominal pain, and fatigue or lethargy in IBD patients, except for nausea and vomiting. No difference in symptom improvement was found in patients with different subtypes, since LFD resulted in similar results of FGS alleviation in both CD and UC patients. The evidence that both IBS-SSS and GSRS scores significantly decreased in patients with LFD intervention further supports the effect of LFD. Water in the small intestinal increases through osmotic potential by absorbing fermentable carbohydrates, such as fructose and mannitol. Intestinal gas (wind) increases through fermenting food by intestinal bacteria, such as fructans and galacto-oligosaccharides [27]. Increased intestinal water and gas appear to play an integral role in triggering symptoms of IBS, such as bloating, abdominal pain, excessive flatus, and altered bowel habit [35]. Approximately 20-60% of IBS patients complained that some food elements trigger their FGSs, especially the 'gas-producing' food (e.g., dairy products, certain fruits, wheat, pulses and legumes, cruciferous vegetables, etc.), with symptoms improved when removing these food items from their diet. Indeed, food hypersensitivity, food allergy, food intolerance, and nonceliac gluten sensitivity are considered to be responsible for these food-related symptoms [36]. In consideration of the proposed mechanism of action of LFD, the top three greatest beneficial symptoms were bloating, flatulence, and borborygmi, which was consistent with previous studies and our meta-analysis in both IBS and IBD [25,35,37,38]. It is noteworthy that no difference was found in nausea or vomiting between groups. The result may be explained by continuous immune activation after LDF intervention, since, compared to other FGSs, nausea or vomiting is the symptom most indicative of the elevated level of interleukin-2 [39].
IBD is a chronic relapsing-remitting gastrointestinal disease. Treatment for IBD consists of diminution or elimination of disease activity and optimization of health-related QoL [40]. The uncertainty of the symptoms and the unpredictability of this clinical condition is highly demanding for IBD patients and deteriorates their QoL [41,42]. Thus, it is obvious that the QoL of patients may be affected by the disease course (extent, severity, and pattern of symptoms' relapse), prescribed therapy (efficacy, side effects, and burden of administration), and psychosocial factors [40,[43][44][45]. IBD patients who suffer from FGSs are more likely to experience anxiety and depression [46]. In the present study, there was a conflicting conclusion of the efficacy of LFD on QoL: a decreased SIBDQ score and an IBS-QoL score with no significant difference. One reason for this may be that IBS-QoL focuses on the impact of stool output, while SIBDQ centers on the multifactorial impact, such as psycho-emotional functioning, systemic symptoms, bowel symptoms, and social functioning. In agreement with this expectation, no beneficial effects were observed in the stool consistency. Thus, it is undeniable that LDF positively affects FGS, and the QoL may be influenced by FGSs in IBD patients.
Stool consistency commonly refers to the rheology or viscosity of the stool, which is strongly dependent on the stool water content [47]. The BSFS is the most widely used scale to quantify stool consistency [48,49]. Diarrhea is the hallmark and the first symptom associated with IBD and appears in 77% of UC patients and 82% of CD patients [50]. The pathogenesis of IBD-associated diarrhea is essentially an outcome of mucosal damage caused by persistent inflammation. Altered expression and/or function of epithelial ion transporters and channels cause electrolyte retention and water accumulation in the intestinal lumen of IBD patients. In addition, aberrant barrier function further contributes to diarrhea via the leak-flux mechanism [51]. Our results do not suggest significant improvement in stool consistency, in terms of both the BSFS score and the number of normal stool consistency (type 3-4 of BSC) after LFD, suggesting no improvement in persistent mucosal inflammation of IBD by LFD.
By comparison, in the aspect of clinical remission, more importance needs to be attached to mucosal healing in IBD management, for the latter predicted a durable complete remission [52]. Several studies have found that both the Crohn's Disease Activity Index (CDAI) and HBi had low specificity and did not correlate well with the endoscopic or histological disease activity of CD patients [53,54]. A considerable proportion of patients who reported clinical remission have mucosal inflammation [55]. The Mayo score is deemed to be more reliable for assessing disease activity of UC patients, for it includes endoscopic score and physicians' clinical assessments, while the CDAI or HBi score attaches more attention to subjective symptoms of CD patients [56]. It is worthy of note that FC is now widely recommended as a sign of intestinal mucosal healing [57]. Low FC has been demonstrated to predict sustained clinical remission in IBD patients [58]. According to our study, pooled data showed a decrease in HBi score and no significant difference in the Mayo score and level of FC. Nevertheless, it should be noted that the study subjects remained in clinical remission in a majority of the included articles. Whether a slight reduction in HBi score in our study can truly reflect a variation in disease activity is yet to be determined. Additionally, it remains to be shown if there is a ceiling effect created by low disease activity on improvement of inflammation by LFD. Hence, further research with a larger sample size and more comprehensive analysis is warranted to validate our results.
The limitations of our study are as follows. Firstly, there were different evaluation standards to assess the relief of FGSs among distinct researchers. With no unified standard, there could be controversial results in different studies, while in the included RCTs, the diets were not standardized and specified in the control group, which might have produced result bias due to the different dietary habits in different regions. Secondly, there was a relatively smaller sample size of the non-RCTs and RCTs included in this meta-analysis. This potentially reduces the reliability of the results. The drawback is more pronounced in subgroup analyses. Additional studies on this topic should be developed to address this question. Thirdly, significant heterogeneity was found among the included studies, which may potentially impact the results of the meta-analysis. The cause of the heterogeneity is still unclear, which may be attributed to the inconsistent research methods and a relatively small number of primary studies. To deal with the potential heterogeneity and provide quality evidence, we performed subgroup analysis according to the statistical method and used a random-effect model suggested by Liberati et al. [59]. Despite the above, it is still the most comprehensive and rigorous meta-analysis to date. Detailed subgroup analysis would benefit basic researchers the most, and no publication bias was found during analysis.
It should be emphasized that despite these improvements in FGSs, intervention using the LFD in IBD should be carefully considered and closely monitored. Indeed, the included studies did not report general or severe adverse effects of the short-term LFD intervention. However, it is important to take into consideration patient adherence and the risk of compromising nutritional status with a long-term restrictive diet. It is well known that undernutrition is common in IBD. Therefore, the use of restrictive diets should be supervised by a dietitian [60], associated with the monitoring of vitamin and mineral deficiencies and proper supplementation accordingly [61].

Conclusions
In conclusion, our meta-analysis demonstrates that LFD has a favorable role in alleviating FGSs in IBD patients, yet without significant benefits in improving stool consistency and mucosal inflammation. Consuming LFD based on professional advice from health care professionals is recommended for IBD patients with problematic FGSs, especially those in remission. Moreover, well-designed and large-scale RCTs are required in the future to confirm the findings and develop the optimal LFD strategy for IBD.