Low FODMAP Diet for Functional Gastrointestinal Symptoms in Quiescent Inflammatory Bowel Disease: A Systematic Review of Randomized Controlled Trials

A low FODMAP diet (LFD) has been hypothesized to relieve symptoms of functional gastrointestinal disorders (FGD) in patients with inflammatory bowel disease (IBD). The aim of the study was to systematically review the literature for randomized controlled trials (RCTs) assessing the effectiveness of the LFD in patients with IBD and FGD. Four databases were searched, but a meta-analysis was not performed due to methodological and outcomes heterogeneity. Four RCTs fulfilled the criteria, with three having some concerns in their risk of bias assessment. All interventions compared the LFDs against a “typical” or sham diet, spanning in duration from 21 days to 6 weeks. Quality of life was improved in two RCTs, while revealing inconsistent findings in the third trial, based on different assessment tools. The fecal assays revealed non-significant findings for most variables (fecal weight, pH, water content, gene count, and gut transit time) and inconsistent findings concerning stool frequency and short-chain fatty acids concentration. Levels of fecal calprotectin, CRP, or T-cell phenotype did not differ between intervention and comparator arms. Two RCTs reported a reduction in abdominal pain, while results concerning pain duration and bloating were inconsistent. In one trial, energy intake was considerably reduced among LFD participants. Regarding gut microbiota, no differences were noted. A considerable degree of methodological and outcome heterogeneity was observed, paired with results inconsistency. The available data are not sufficient to justify the claim that an LFD induces relief of FGD symptoms, although it may pave the way to a placebo response.


Introduction
The form and nutrient content of ingested food may trigger a variety of gastrointestinal (GI) symptoms through a matrix of different mechanisms, including bacterial fermentation altering gut microbiota, the induction of distinct osmotic load effects in the small bowel and colon, the production of gas in the GI tract, and the activation or suppression of immune responses [1,2]. Putative anti-inflammatory foods and elimination diets, including the low fermentable oligosaccharides, disaccharides, monosaccharides, and polyols (FODMAPs) diet (LFD), have been proposed as complementary regimes alleviating symptoms of functional gastrointestinal disorders (FGD) [3][4][5][6]. Since the LFD was designed [7], its adoption has gained traction and today, it is often recommended in clinical practice [8], with evidence synthesis indicating that adoption of the LFD reduces symptoms of irritable bowel syndrome (IBS) and functional GI symptoms (FGS) in general [9][10][11][12].
FGS, however, are also frequent among patients with inflammatory bowel diseases (IBD). It has been estimated that at least one-third of quiescent Crohn's disease (CD) and ulcerative colitis (UC) patients experience functional intestinal problems including abdominal bloating, flatulence, pain, and changes in stool consistency and frequency unrelated to concurrent inflammation [13]. The overlap of IBD and functional gastrointestinal disorders (FGD) is often associated with lower quality of life (QoL) [14] and increased anxiety and depression [15]. Although a few studies have examined the LFD among IBD patients with FGD, the subsequent guideline recommendations advocating the adoption of an LFD as a possible treatment appear weak and of low evidence [16][17][18]. One meta-analysis [19] of two randomized controlled trials (RCTs) and an equal number of case-control studies supported use of the LFD; however, since its publication, more research has emerged, the results of which have not yet been integrated into any synthesis of evidence.
Thus, the aim of the present systematic review was to update the evidence evaluating the effectiveness of adherence to an LFD in relieving functional gastrointestinal disorders among IBD patients in remission who are experiencing FGS.

PICO and Research Question
The research question of the study was: What is the effect of adopting an LFD among patients with quiescent IBD who have IBS or other functional GI symptoms? The PICO criteria were: Patients with IBD in remission experiencing FGS (P); the Intervention (I) was adherence to an LFD; the Comparator (C) was a standard/sham diet, placebo, or non-FODMAP elimination diet; and the Outcome (O) was a change in FGS. The protocol was registered at the Center for Open Science (OSF) (osf.io/6qv3z) and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [20] guidelines were adhered to.

Search Strategy
Three researchers (K.G., M.P.N., and M.G.G.) independently reviewed the literature systematically, searching on the PubMed (https://pubmed.ncbi.nlm.nih.gov/), Cochrane Central Register of Controlled Trials (CENTRAL; https://www.cochranelibrary.com/central), Scopus (https://www.scopus.com/home. uri), and Clinical.Trials.gov (https://www.clinicaltrials.gov/) databases for eligible randomized controlled trials (RCTs). In cases of lack of agreement, a senior researcher (D.G.G.) solved the issues through discussion and thorough appraisal of the RCTs. Search terms with a combination of MeSH terms (whenever applicable in each database) included: (inflammatory bowel disease), (IBD), (Crohn's disease), (ulcerative colitis), and (FODMAP). The search ended in April 2020, without any restriction being imposed on the language of the retrieved trials. A detailed search strategy for PubMed is presented in Figure 1.

Inclusion and Exclusion Criteria
Studies were included in the qualitative synthesis when: (1) there was an RCT design (with no restriction on the randomization method or the design of the intervention arm); (2) there was a comparison of at least one treatment using a FODMAP elimination diet to a non-FODMAP elimination scheme, sham or usual diet, or placebo; (3) the study included patients with IBD in remission; (4) where an FGD was diagnosed; (5) without any restrictions concerning the age of participants or (6) the intervention duration; (7) published at any date until April 2020.
Nutrients 2020, 12, x FOR PEER REVIEW 3 of 22 comparison of at least one treatment using a FODMAP elimination diet to a non-FODMAP elimination scheme, sham or usual diet, or placebo; (3) the study included patients with IBD in remission; (4) where an FGD was diagnosed; (5) without any restrictions concerning the age of participants or (6) the intervention duration; (7) published at any date until April 2020. Exclusion criteria were: (1) studies without an RCT design; (2) studies performed on animals or (3) on patients with other diseases as well as IBD; (4) studies which included IBD patients who were not in remission or (5) did not have FGD; (6) studies using interventions other than a LFD; (7) studies comparing a FODMAP elimination diet to other FODMAP elimination schemes (less/more intense or different constituents); (8) studies lacking a comparator or (9) using a one group cross-over design.

Outcomes of Interest
All documented outcomes were considered as important, including changes in QoL, analysis of microbiota, IBD and FGD severity and symptom scores, dietary intake, and markers of immunity and inflammation.

Risk of Bias and Quality of the RCTs
The Cochrane Risk of Bias (RoB) 2.0 tool [21] was used to assess bias in the retrieved RCTs, with judgments falling in the categories of "low risk of bias", "some concerns", or "high risk of bias". Two independent researchers (S.T.P. and K.G.) assessed RoB and a senior researcher (D.G.G.) intervened when an agreement was not reached. In parallel, the quality of the included studies was evaluated by two reviewers (M.P.N. and K.G.) using the Oxford quality score [22].

Data Extraction
Three researchers (M.G.G., K.G., S.T.P.) independently extracted data into predefined Excel © forms. The extraction involved characteristics of the trials (e.g., origin, design, masking, ethics, funding, analysis performed), details concerning the participants (e.g., n in each stage, IBD diagnosis, IBS/FGD diagnosis, age), nature of the intervention (e.g., intervention, comparator, intervention duration, wash-out duration, compliance assessment, drop-outs, adverse events), and outcomes of interest. Exclusion criteria were: (1) studies without an RCT design; (2) studies performed on animals or (3) on patients with other diseases as well as IBD; (4) studies which included IBD patients who were not in remission or (5) did not have FGD; (6) studies using interventions other than a LFD; (7) studies comparing a FODMAP elimination diet to other FODMAP elimination schemes (less/more intense or different constituents); (8) studies lacking a comparator or (9) using a one group cross-over design.

Outcomes of Interest
All documented outcomes were considered as important, including changes in QoL, analysis of microbiota, IBD and FGD severity and symptom scores, dietary intake, and markers of immunity and inflammation.

Risk of Bias and Quality of the RCTs
The Cochrane Risk of Bias (RoB) 2.0 tool [21] was used to assess bias in the retrieved RCTs, with judgments falling in the categories of "low risk of bias", "some concerns", or "high risk of bias". Two independent researchers (S.T.P. and K.G.) assessed RoB and a senior researcher (D.G.G.) intervened when an agreement was not reached. In parallel, the quality of the included studies was evaluated by two reviewers (M.P.N. and K.G.) using the Oxford quality score [22].

Data Synthesis
A meta-analysis was intended for quantitative outcomes, where a minimum of at least three studies recording the outcome of interest would be apparent. Categorical outcomes were intended to be synthesized with odds ratios, along with the 95% confidence intervals (CI), while outcomes with continuous variables were intended to be expressed as mean differences (or standardized mean differences) with their corresponding 95% CI. A random-effects meta-analytic model was initially chosen, and statistical heterogeneity was scheduled for investigation. Results of the meta-analysis were intended to be summarized visually using forest plots, and publication bias would be assessed with funnel plots. The significance level of the meta-analysis was set at alpha = 0.05.
the Pedersen et al. RCT was implemented in Denmark [27][28][29]. The trial of Pedersen et al. [27][28][29] had open-label masking, and the Bodini [26] and Cox [24,25] RCTs were single-blind. The trial of Halmos et al. [23] reported only participant blinding in the manuscript text (single-blind), while referencing a previous study performed using double-blind masking [30], although the principle diagnoses of participants between the two studies failed to match. Only the Halmos et al. [23] RCT adopted a crossover design and the remaining trials compared parallel interventions [24][25][26][27][28][29].   Full-text (n = 1) Full-text (n = 1) and poster abstract (n = 1) Full-text (n = 1) Full-text (n = 2) and poster abstract (n = 1) √ included in the study; * Defined by PGA, stable medications, no IBD flare in the previous 6 months, fCAL < 250 µg/g, and serum CRP < 10 mg/L; oe Defined by HBi < 5; † Median, range; ƒ Median, interquartile range; § Masking was reported as double-blind through a reference to a previous protocol [30], however (a) participants did not match between the two publications [23,30], and (b) in the latter publication [23], blinding of investigators was not reported in the manuscript text; š No information on IBS/FGS was provided in the manuscript presenting the RCT's results [23]. A reference to a previous publication [30] is provided for more information concerning the sample; however, a mismatch in the sample is evident, as in the first reference, participants included IBS patients only [30], excluding all with diagnosed IBD, whereas in the latter publication [23], IBD was the primary inclusion criterion.
Nutrients 2020, 12, 3648 8 of 22 The number of included participants ranged from as few as 9 in the Halmos trial [23] to a total of 89 recruited by Pedersen [27][28][29] for adult patients with quiescent IBD and FGS. Halmos et al. [23] were the only investigators to limit their work to patients with CD, whereas the other studies also included some patients with UC. Cox et al. [24,25] presented the differential FGD diagnosis of all participants (functional bloating/diarrhea, mixed/unsubtyped IBS, or diarrhea-predominant IBS). Bodini and colleagues [26] reported only on patients with IBD-IBS. Petersen and associates [27][28][29] recruited patients with IBD and FGS, whereas in the Halmos et al. RCT [23], no reference to the FGD status of participants was made in the methodology, but rather, it was implied in their introduction. All participants had an FGD diagnosis based on the Rome III [32] or IV [31] criteria.
The FODMAP content of each diet was reported to have been assessed through the FODMAP food composition database (Monash University, Melbourne, Australia) or via food analysis using high-performance liquid chromatography and enzymatic assays. In the Halmos trial, LFD was determined as ≤0.5 g of FODMAP per sitting [37]. Compliance was mostly assessed by dietitians [23,26,30], using food-frequency questionnaires (FFQs) [27][28][29] or food diaries [24][25][26]. Cox and associates [24,25] additionally used one question at the end of the intervention to assess adherence to the LFD "'During the 4-week trial I have followed the diet . . . ': never/rarely (<25% of the time), sometimes (25-50% of the time), frequently (51-75% of the time) or always (76-100% of the time)". The Halmos trial [23] was the only one where the fiber content of the LFD was adjusted to that of the typical control diet using supplemental psyllium and resistant starch.

Outcomes of Interest
Disease activity was determined using a variety of tools, including the IBD Control questionnaire (IBD-Control-Q) [38], the Harvey Bradshaw Index (HBI) [33] for patients with CD, and the Partial Mayo Score PMS [36] or the Simple Clinical Colitis Activity Index (SCCAI) [39] for those with UC (Table 2).
Stool samples were collected in all four RCTs, and fecal calprotectin (fCAL) was assessed using an enzyme-linked immunosorbent assay (ELISA). In the Cox [24,25] and Halmos [40] trials, microbiome composition was also assessed using quantitative metagenomic pipeline and polymerase chain reaction (PCR) on DNA fecal samples, respectively. In parallel, different stool indices were recorded in each trial, including consistency via the Bristol Stool Form Scale (BSFS) [41], fecal water content (FWC), stool frequency, weight, pH, and short-chain fatty acid (SCFA) concentration, and gut transit time ( Table 2).

Risk of Bias of the Included RCTs
Summary of the risk of bias in the included RCTs revealed (Figure 3) that the trial by Cox et al. [24,25] was of low risk in all examined domains. On the other hand, the studies by Bodini et al. [26] and Halmos et al. [23] had some concerns regarding the randomization process as well as the overall bias. The Halmos et al. [23] trial also raised some concerns regarding missing outcome data. Some concerns were raised regarding missing outcome data in the trial conducted by Halmos et al. [23]. In the Bodini et al. [26] RCT, concerns were raised regarding the measurement of the outcomes and the selection of the reported findings. Changes in the dietary intake of participants following the interventions were assessed by Halmos and Cox [23][24][25], using FFQs [45] and food diaries ( Table 2).

Risk of Bias of the Included RCTs
Summary of the risk of bias in the included RCTs revealed (Figure 3) that the trial by Cox et al. [24,25] was of low risk in all examined domains. On the other hand, the studies by Bodini et al. [26] and Halmos et al. [23] had some concerns regarding the randomization process as well as the overall bias. The Halmos et al. [23] trial also raised some concerns regarding missing outcome data. Some concerns were raised regarding missing outcome data in the trial conducted by Halmos et al. [23]. In the Bodini et al. [26] RCT, concerns were raised regarding the measurement of the outcomes and the selection of the reported findings.

Fecal Assays
Analyses of fecal samples in the Halmos and Cox RCTs [23][24][25] revealed inconsistent findings concerning stool frequency and SCFA concentration and a lack of differences in stool pH. Individual trials revealed that the LFD did not induce any significant effects on the FWC, fecal weight, gene count, or gut transit time [23], or in the proportion of produced stools with normal consistency [24,25].

Fecal Assays
Analyses of fecal samples in the Halmos and Cox RCTs [23][24][25] revealed inconsistent findings concerning stool frequency and SCFA concentration and a lack of differences in stool pH. Individual trials revealed that the LFD did not induce any significant effects on the FWC, fecal weight, gene count, or gut transit time [23], or in the proportion of produced stools with normal consistency [24,25]. Concerning fCAL, three trials (Halmos, Cox, and Pedersen) reported no significant effect [23][24][25][27][28][29], but Bodini et al. [26] noted a reduction following the LFD (Table 3).

Markers of Inflammation and Immunity
According to Bodini, Cox, and Pedersen, after commencing the LFD intervention, no difference was demonstrated in the CRP levels of participants, as compared to the controls [24][25][26][27][28][29] (Table 3). The peripheral T-cell phenotype was also largely unchanged according to Cox [24,25], with only the number of α4β7 positive Vδ2 T-cells being reduced post intervention.

Relief of FGD Symptoms
Findings concerning FGD symptoms were heterogeneous in terms of assessment tools and generally inconsistent when more than one trial was compared and of low evidence quality, as the majority of data points were unique to individual trials ( Table 3). The severity of FGD symptoms improved in the Halmos and Pedersen trials [23,[27][28][29] as measured with a VAS or the IBS-SSS, although Cox [24,25] failed to demonstrate any improvement using the latter tool. Reduction in abdominal pain was also reported by the Halmos and Pedersen RCTs [23,[27][28][29] using different tools. However, when the analysis was based on a common tool-the IBS-SSS-Cox and Pedersen [24,25,[27][28][29] demonstrated inconsistent findings regarding bloating symptoms and pain duration, although both showed a reduction in stool frequency and improved consistency. Finally, using the GSRS, Cox et al. [24,25] reported reduced flatulence symptoms among LFD participants at the end of the trial.

Changes in the Dietary Intake Following an LFD
Adherence to an LFD induced inconsistent findings in the four RCTs (Table 3). In the Cox et al. trial [24,25], equivalent energy intake between the intervention and the control group was not monitored, and the use of oral nutrient supplements (ONS) was forbidden. This resulted in reduced energy, protein, total fat, sugar, calcium, iodine, and phosphorus intake in the LFD participants. On the other hand, in the Halmos et al. trial [23], participants were instructed to eat according to their appetite to fulfill all Australian food group-specific recommendations, with their patients demonstrating an average daily energy intake of approximately 8 MJ (1900 kcal). In parallel, fiber intake was increased in the LFD group through the provision of ONS. These factors synergistically reduced possible adverse LFD effects concerning the nutrient/food group adequacy of participants, as with the exception of dietary starch, analysis of quantiles of food groups intake and comparison to the Recommended Daily Allowances failed to record differences in the intake between the LFD and control groups [23].  [46]; HBi, Harvey Bradshaw Index for CD [33]; HR-QoL, IBS Health-related Quality of Life [43]; IBD, inflammatory bowel diseases; IBD-Q, inflammatory bowel disease-Quality of life [34,35]; IBS, irritable bowel syndrome; IBS-QoL, IBS Quality of Life [44]; IBS-SSS, IBS Severity Scoring System [49]; LFD, low FODMAP diet; NR, not reported; NS, not significant between intervention groups; PMS, Partial Mayo Score for UC [36]; SCCAI, Simple clinical colitis activity index [39]; SCFA, short-chain fatty acids; SF36, short-form 36 [42]; SIBDQ, Short Inflammatory Bowel Disease Questionnaire; UC, ulcerative colitis; VAS, Visual Analogue Scale; ↓ lower in LFD group as compared to controls at the end of the trial; ↑ higher in LFD compared to controls at the end of the trial; * psyllium (3 g/day) and resistant starch (5 g Hi-Maize 220/day) were added to the LFD diet (National Starch and Chemical Company, Bridgewater, NJ, USA), respectively, to ensure that only the FODMAP content of the two diets differed; † participants were advised to eat to their appetite with daily energy intake averaging at approximately 8 MJ. Table 4 details all changes in fecal bacteria abundance following the LFD intervention compared with the control diet. No differences were noted in the total species abundance (relative or absolute) or the total Bifidobacteria sp. in either trial where this was studied [23][24][25]. Cox et al. [24,25] showed no differences in the gene count, phyla distribution, α and β-diversity, nor in targeted Bifidobacteria including animalis, bifidum, breve, and pseudocatenulatum species. On the other hand, a reduction in the relative abundance of Bifidobacterium adolescentis, dentium, and longum was recorded in the LFD group compared with the controls [24,25]. The Halmos and Cox RCTs [23][24][25] were unable to reach a unanimous finding concerning the total Faecalibacterium prausnitzii abundance post LFD intervention. Finally, Halmos and associates [23] suggested a reduction in the absolute and relative fecal content of Clostridium cluster XIVa, Akkermansia muciniphila, and the relative abundance of Ruminococcus torques following the LFD, and a lack of significant difference concerning Roseburia, Lactobacilli sp., Ruminococcus gnavus, and Clostridium cluster IV. IBD, inflammatory bowel diseases; LFD, low FODMAP diet; NS, not significant; RCT, randomized controlled trial; ↓ lower in the LDF participants as compared to the controls; ↑ higher in the LFD participants as compared to the controls; * the comparator was a sham diet; † the comparator was the typical Australian diet.

Data Synthesis
Considering the methodological and clinical heterogeneity observed in the individual studies, a meaningful meta-analysis with respect to any quantitative outcome from at least three RCTs was not considered feasible, as it would have resulted in unreasonably excessive variation in any attempt to synthesize the data. Table 5 details all ongoing RCTs registered in Clinicaltrials.gov, examining the efficacy of the LFD on patients with IBD. Three trials in total, implemented in Mexico (NCT04143633), Iran (NCT03644602), and Denmark (NCT02469220), have results pending to be published. All trials have both subjective and objective outcomes, with intervention duration ranging between 4 and 10 weeks.

Discussion
Careful inspection revealed that the RCTs were highly heterogeneous in terms of design, participants, and outcomes. Accordingly, any attempt to recommend adherence to the LFD for relief from FGS in patients with IBD will be based on inadequate evidence. The four retrieved RCTs yielded inconsistent findings concerning all outcome domains, including disease severity, QoL, FGS relief, gut microbiota, nutrient intake, immunity and inflammation markers, stool characteristics, and fecal composition.
According to the included RCTs, conflicting results were produced concerning disease severity and FGS relief following adherence to the LFD. The pathophysiology of IBD is well understood, with the presence of chronic mucosal inflammation paired with a dysregulated immune response due to type 1 T-helper cells (in CD) or type 2 T-helper cell (in UC), resulting in partial loss of bowel integrity, while impeding epithelial barrier regeneration [50]. In FGD and IBS in particular, a persistent immune activation is apparent; therefore, even in remission, patients with IBD-IBS demonstrate dysbiosis [51], chronic relapse phases [52], a compromised immune response, an increased gut permeability, and a discorded brain-gut axis, all of which are augmented as compared to patients with IBD alone. With an overlap in signs, symptoms, pathophysiology and genetic profile in most cases, the two groups of conditions (the IBDs and the FGDs) are sometimes even regarded as the evolution of a similar disease [53]. However, even though the LFD is currently appraised as a possible treatment for IBS, the differences might well be too great for any useful extrapolation to patients with IBD-IBS who tend to exhibit more severe clinical pathophysiology with many more objective phenomena and who perhaps develop surges of the FBD elements during periods of increased stress and anxiety [15].
According to the literature, FODMAP intake has a direct effect on the gut microbiota, and adherence to an LFD reduces luminal bifidobacteria [10,54,55], resulting in the loss of prebiosis [56,57]. As far as patients with IBS are concerned, they tend to exhibit a greater degree of instability and a diminished microbiota diversity, both of which remain unchanged following adherence to an LFD [54,58,59]. In the present review, a lack of agreement was noted between the Halmos and Cox RCTs [23][24][25] concerning the effect of LFD on targeted bacteria. A difference in the severity of pathophysiology among the recruited participants might, in part, explain this result. Cox and associates [24,25] used patients with IBD and mixed FGD diagnoses, whereas Halmos et al. [23] selected CD patients with FGS, although detailed characteristics of the symptoms or FGD diagnosis were not provided, other than in reference to a previous article using patients with IBS only [30]. Undoubtedly, the two samples were highly heterogeneous, which might have induced different responses to the LFD. In parallel, according to Simrén [60], it is not the altered gut microbiota that induces FGS relief following LFD adherence, but rather a synergy of the FODMAP composition of the meals and the diet-microbiota produced interactions. This is why LFD studies often fail to produce significant differences in the gut microbiota from the control groups, and why GI symptoms often reappear immediately after discontinuation of the LFD [60].
Conflicting findings were also noted concerning the QoL of patients after adhering to the LFD. Considering that the IBD diet is already restrictive to some degree, and that restrictive diets can sometimes be stressful for patients with chronic disease, any effort to eliminate more foods, or impose further dietary restrictions might hamper the adherence rate, produce opposite results, and have a negative effect on the health and QoL of patients [61]. In the LFD, in particular, available dietary choices are restricted to a great degree, reducing long-term adherence [59,62]. Ooi [63] and Halmos [40] noted that extensive or inappropriate use of the LFD could have a negative impact on the health of patients. On the other hand, the duration of most LFD trials is limited and cannot ensure long-term efficacy comparable to the drug trials [64].
In parallel, due to its restrictive nature, long-term adherence to the LFD might compromise nutrient status [56,65], posing an additional risk for malnutrition among patients with IBD. Although most IBD diets fail to provide adequate amounts of all nutrients [66], the LFD in particular results in reduced fiber consumption. Among the RCTs included herein, Cox et al. [24,25] demonstrated a non-significant difference in fiber consumption, paired with a reduced energy intake in the LFD group, averaging 1697 kcal/day as compared to 1918 kcal/day demonstrated by the controls at the end of the trial. This energy intake amount is relatively low, and possibly indicative of a great proportion of low-energy reporters in the sample and/or diet records of low accuracy. Considering that in dietary analysis, the intake of all nutrients is depended on the recorded energy intake, low accuracy in the energy records might inevitably lead to inaccurate fiber consumption estimation. On the other hand, Halmos [23] accounted for the low fiber content of the LFD in advance, providing all participants with ONS, which resulted in a non-significant fiber consumption between the two groups. The reported low fiber intake when adhering to the LFD and possibly, lower energy consumption, is an important note to consider, especially in patients with IBD. The latter often fluctuate on the verge of under-nutrition and nutritional deficiencies [56], with the majority failing to receive dietetics consultation related to IBD [67]. Other concerns raised include the possible low vitamin D levels as a result of reduced dairy intake, although this was not demonstrated in the present review [59]. To ensure nutritional adequacy, Halmos [40] suggested a more gentle LFD as compared to patients with IBS alone; however, the effects of less restrictive LFDs on reducing FGD symptoms have never actually been assessed. The FODMAP inventors recommend a FODMAP-reintroduction diet in intervals lasting for two months to ensure nutrient adequacy [7]. Nevertheless, a follow-up of patients with IBD and IBS revealed that compliance was a difficult task, with only a third adhering to the LFD for more than 18 months, indicating that long-term elimination diets might be difficult to follow in the long run [68].
Undoubtedly, all nutrition interventions in IBD carry a variety of limitations. In nutrition research, expectations concerning the consumption of particular foods, personal beliefs, prior dietary advice received, dietitian reassurance, and sensory preferences, all act synergistically in creating a placebo response [69]. On the other hand, in IBS research, high rates of both placebo and nocebo responses have been suggested to occur [70]. Although in IBD alone, objective outcomes can be evaluated to limit this placebo response phenomenon [71], in the case of IBD-IBS, the placebo effect appears to be quite profound. This is augmented by a possible unmasking of the diet therapy, the discordant brain-gut axis exhibited in both IBD and FGD, and the fact that FGD severity is greatly dependent on the stress levels [72,73] and underlying biopsychosocial pathogenesis [31]. Therefore, adhering to a diet regime that is considered "healthy" might reduce anxiety and subsequently, alleviate IBS symptoms, creating a placebo response.
In parallel, examining the symptomatic effects of a diet entails substantial challenges in terms of both trial design and implementation [64]. This is why the use of subjective measures evaluating symptom severity often leads to great inconsistency in the interpretation of IBS symptoms [74]. To correct this, composite outcome scores have been proposed as optimal endpoints in IBS research [75], although these were not employed in any RCT included herein. Most RCTs failed to include objective outcomes like immune activation markers, changes in the gut microbiota or on the gut lumen [1] and interestingly, when objective outcomes were assessed, such as the CRP, fCAL, or T-cell phenotype, the lack of significant differences post-LFD adherence was apparent.
Concerns have also been raised regarding the appropriate comparator for examining the efficacy of the LFD [59]. Comparison to the usual diet, as performed in the Halmos, Bodini, and Pedersen trials [23,[26][27][28][29], does not prove that the LFD is superior to the conventional IBS medical nutrition therapy (MNT). For this reason, studies comparing the LFD to other dietary treatments for IBS failed to produce significant results [76][77][78]. Therefore, by design, the existing RCTs appear to be flattering the LFD, without actually examining its effectiveness, or controlling for the subjectivity of the selected outcomes, by using composite endpoints.

Conclusions
Without a doubt, diet therapies are the most challenging to study to produce high-quality evidence [79,80]. Nevertheless, they are sought by patients and can often provide effective non-pharmacological solutions with limited adverse events. Given the known psychological effects of diet [69], it is possible that the LFD might induce a placebo response in some patients which is highly desirable. However, through careful critical appraisal of the evidence, the present systematic review failed to provide adequate evidence in terms of quality and quantity to support recommendations for an LFD for IBD patients with FGD. Publication of the results from the three ongoing RCTs is expected to add more weight to the evidence examining the efficacy of the LFD in patients with IBD and FGD. Nevertheless, as with every diet therapy, expert guidance and personalized support with FODMAP-experienced dietitians might help avoid nutritional inadequacies, while maintaining long-term adherence [81,82].