Predefined Diets in Patients with Inflammatory Bowel Disease: Systematic Review and Meta-Analysis

Inflammatory bowel disease (IBD) is a chronic disease mediated by the immune system and characterized by the importance of diet in pathological development. This study aims to understand how the use of predefined diets can affect the adult population diagnosed with IBD. We conducted a systematic review and meta-analysis. From the different databases (MEDLINE, Scopus, Cochrane, LILACS, CINAHL, and WOS), we found 4195 registers. After a review process, only 31 research studies were selected for qualitative synthesis and 10 were selected for meta-analysis. The variables used were Crohn’s Disease Activity Index (CDAI) for patients with Crohn’s Disease (CD) and fecal calprotectin (FC), C-Reactive Protein (CRP), and albumin (ALB) for patients with IBD. Predefined diets have been shown to have partial efficacy for the treatment of IBD and are compatible with other medical treatments. CDAI improved but with reasonable doubts due to the high heterogeneity of the data, while no differences were observed for ALB, FC, and CRP. More studies that evaluate the influence of predefined diets on IBD patients are needed due to the great variability in diets and the tools used to measure their effects.


Introduction
Inflammatory bowel disease (IBD) is characterized by chronic inflammation of the gastrointestinal tract with periods of remission or recurrence and includes both Crohn's disease (CD) and ulcerative colitis (UC) [1].
In CD, inflammation can be present in any area of the entire gastrointestinal tract, whereas in UC, the inflammatory process affects only the colon [2].
The symptoms of this type of disease are diverse, including the appearance of diarrhea, bloody stools, abdominal pain, fatigue, weight loss, etc. [3]. The prevalence exceeds 0.3% in North America, Oceania, and many European countries, and the incidence of these types of pathologies has increased rapidly in recently industrialized countries, producing a high burden on health systems [4,5].
IBD, apart from being treated with expensive medical treatments to alleviate its activity, leads to a decrease in the patient's quality of life that can affect the degree of disability and work productivity and is associated with more symptoms of anxiety and depression [6][7][8].
In addition, malnutrition and specific nutritional deficiencies are frequent among these types of patients, depending on the state and/or progress of the disease, and most patients impose dietary restrictions based on their own beliefs [9,10].
The etiology of IBD is unknown; in fact, it is a multifactorial disease. However, Westernised lifestyles and diets are one of the main drivers of the increasing incidence [11,12].

Materials and Methods
To achieve this objective, a systematic review was conducted in agreement with the procedures and verification list described by PRISMA [28]. Afterwards, a meta-analysis on the more common results was conducted.

Systematic Review
A search of scientific works was conducted in the MEDLINE database, through the open retrieval system on the Internet such as PubMed, the Cochrane Library, Scopus, Web of Science, CINAHL, and LILACS. Studies conducted up to 10 January 2020 were compiled.

Inclusion and Exclusion Criteria
The studies selected had to comply with the following inclusion criteria: refers to an adult population (older than 18) diagnosed with IBD; studies the effect of predefined diets within IBD; be clinical trials and observational studies; be in English, Spanish, Portuguese, French, or German.
The following articles were excluded: those that referred to the infant population, to animals, or to the use of predefined diets in a healthy adult population and those that sought the effect of specific foods or nutrients in IBD, without a clear diet designation, that were case report studies, or that were based on secondary sources.

Search Equation
To include content linked to the intervention and predefined diets, two specific descriptors were used, such as "Nutrition Therapy" and "Diet", and the same terms were used in the title or abstract.
For content linked to the population, we utilized the descriptor that referred to the disease, "Inflammatory bowel diseases", and its equivalent term in the title or abstract.
Also, the filters "Humans" and "Adult" were utilized to achieve our objective. Therefore, the main search equation designed for this study was as follows: The search equation was adapted to each and all of the databases described previously. The process was conducted in the period from December 2019 to January 2020.

Selection Process
After eliminating duplicate records, the process of selection was conducted in two phases. The first consisted of reviewing the titles and abstracts of all the article records resulting from the adapted search equations and shown by the databases by using the inclusion and exclusion criteria and the objective of the study as the screening measures. The screening and selection of the records/articles were conducted independently by the two researchers, both experts in the fields of nutrition. These researchers agreed on the discrepancies found in order to define the final suitability of the records/articles found in the databases. The precision of the search was calculated based on the ratio of full-text articles selected for the review divided by the number of records found by the search equation and multiplied by one hundred.
The second phase involved the application of inclusion/exclusion criteria to the complete text of all the scientific studies selected in the first phase, thus ensuring the relevance of each one of them. In order to obtain studies that were not accessible via the Internet, we used three methods: Researchgate, the corresponding author, and interlibrary loan.

Evaluation of the Quality of the Studies
Evaluation of the methodological quality of the included studies was performed by two independent researchers, using the CONSORT (Consolidated Standards of Reporting Trials) guide for clinical trials and STROBE (Strengthening the Reporting of Observational studies in Epidemiology) for observational studies.
These guides contain a list of 25 (CONSORT) and 22 (STROBE) essential aspects that should be described in the publication of these studies. For each selected study, one point was assigned for each item present (if not applicable, it was not scored). When an item was composed of several points, these were evaluated independently, giving the same value to each of them, and subsequently an average was made (being the final result of that item), so that in no case could it beat the score of one point per item [29][30][31].

Meta-Analysis
To calculate the effect size of the enteral nutrition on the variables Crohn's Disease Activity Index (CDAI), albumin (ALB), Fecal Calprotectin (FC), and C-Reactive Protein (CRP), a meta-analysis was performed. For this, the model of fixed effects and the model of random effects were used. The results are presented as a forest plot along with the percent heterogeneity and its confidence interval at 95%, the t-value, and the heterogeneity test.
To explore the influence of each study over effect size, we used a leave-one-out method; pooled estimates were calculated omitting one study at a time. In addition, we plotted a scatterplot introduced by Baujat et al. [32]. On the x-axis, the contribution of each study to the overall heterogeneity statistic was plotted. On the y-axis, the standardized difference of the overall treatment effect with and without each study was plotted; this quantity describes the influence of each study on the overall treatment effect. Therefore, studies that fall on the top right quadrant of the Baujat plot have the most influence.
Publication bias occurs only when favorable results are published, and this could have consequences on the results of the meta-analyses if these are included. To analyze the publication bias, a nonparametric analysis was conducted, as proposed by Duval and Tweedie [33] based on the funnel-plot, by estimating and adjusting for the number and outcomes of missing studies in the meta-analysis. Another less-conservative proposal to estimate the number and outcomes of missing studies is the proposal by Copas et al. [34].
A meta-regression could not be performed due to the low number of studies. All calculations were performed within an R programming environment utilizing the packages meta version 4.15-1 [35] and metasens version 0.5-0 [36].

Systematic Review
As a result of the specific search equations used in the different databases, a total of 5645 records of scientific articles were found. A total of 1450 records were duplicated, leaving a total of 4195 records without duplication. In the first phase of the study, exactly 4135 study records were discarded, leaving 60 full-text studies to review, so that the accuracy was 1%. As shown in Figure 1, 2514 records did not study the effect of predefined diets, 576 did not refer to humans, 499 showed that the study utilized a design that was not adequate, 283 did not use an adult population, 189 did not refer to IBD, 64 were still being conducted without showing results, and 10 were written in another language other than the ones cited above.

3.1.Systematic Review
As a result of the specific search equations used in the different databases, a total of 5645 records of scientific articles were found. A total of 1450 records were duplicated, leaving a total of 4195 records without duplication. In the first phase of the study, exactly 4135 study records were discarded, leaving 60 full-text studies to review, so that the accuracy was 1%. As shown in Figure 1, 2514 records did not study the effect of predefined diets, 576 did not refer to humans, 499 showed that the study utilized a design that was not adequate, 283 did not use an adult population, 189 did not refer to IBD, 64 were still being conducted without showing results, and 10 were written in another language other than the ones cited above. In the second phase, 29 studies were removed: 20 because they did not investigate the effects of predefined diets, 6 due to defects in design, and 3 because the patients studied were not adults. Therefore, only 31 research studies were selected, as shown in Figure 1.
In addition, 14 of the studies found showed results that specifically referred to CD, 4 studies referred to UC, and 13 studies showed results for both UC and CD under the category of IBD. Also, 8 studies mentioned the results of the disease in its active form, 3 studies reported disease outcomes of patients under surgery, 15 studies used a population with IBD in remission, and 5 studies did not indicate disease status. Figure 2 shows this information in a chronological manner. In the second phase, 29 studies were removed: 20 because they did not investigate the effects of predefined diets, 6 due to defects in design, and 3 because the patients studied were not adults. Therefore, only 31 research studies  were selected, as shown in Figure 1.
In addition, 14 of the studies found showed results that specifically referred to CD, 4 studies referred to UC, and 13 studies showed results for both UC and CD under the category of IBD. Also, 8 studies mentioned the results of the disease in its active form, 3 studies reported disease outcomes of patients under surgery, 15 studies used a population with IBD in remission, and 5 studies did not indicate disease status. Figure 2 shows this information in a chronological manner. As for the variety of predefined diets used in the studies, a total of 17 different types were found as shown in the Table 1. The total population analyzed in the research studies found included a total of 5331 individuals with IBD: 829 diagnosed with CD and 422 with UC.
The main tools utilized by the researchers to obtain results were scores, biomarkers, and tests to measure the activity of the disease: the Crohn's Disease Activity Index (CDAI), the Harvey-Bradshaw Index (HBI), the Van Hees index (VHI), the Modified Truelove and Witts activity index (MTWAI), the Mayo score (MS), the partial Mayo score (PMS), irritable bowel syndrome severity score system (IBS-SSS), Copenhagen IBS disease courses (CIBSC) visual analogue scales (VAS); biomarkers such as CRP, ESR, the white blood cell count (WBC), levels of albumin (ALB), pre-albumin (PA), transferrin (TRF), hemoglobin, platelet count (PL), alkaline phosphatase (ALP), etc.; and medical tests such as an ileocolonoscopy. Complementary tests were also included, such as urine, feces samples, and Bristol stool (BS) tests. Tests that measured the body's composition were also found, such as anthropometries and bioimpedance, to obtain parameters such as body weight (BW) and body mass index (BMI). Quality of life questionnaires included the IBD Questionnaire (IBDQ), the short IBD Questionnaire (SIBDQ), the United Kingdom version of IBDQ (IBDQ-UK), and the irritable bowel syndrome quality of life questionnaire (IBS-QOL). Table 2 schematically shows the main results found in the selected articles. Table 3 and Table 4 show the scores obtained by the studies for their methodological quality according to the CONSORT and STROBE guidelines. As for the variety of predefined diets used in the studies, a total of 17 different types were found as shown in the Table 1.
The total population analyzed in the research studies found included a total of 5331 individuals with IBD: 829 diagnosed with CD and 422 with UC.
The main tools utilized by the researchers to obtain results were scores, biomarkers, and tests to measure the activity of the disease: the Crohn's Disease Activity Index (CDAI), the Harvey-Bradshaw Index (HBI), the Van Hees index (VHI), the Modified Truelove and Witts activity index (MTWAI), the Mayo score (MS), the partial Mayo score (PMS), irritable bowel syndrome severity score system (IBS-SSS), Copenhagen IBS disease courses (CIBSC) visual analogue scales (VAS); biomarkers such as CRP, ESR, the white blood cell count (WBC), levels of albumin (ALB), pre-albumin (PA), transferrin (TRF), hemoglobin, platelet count (PL), alkaline phosphatase (ALP), etc.; and medical tests such as an ileocolonoscopy. Complementary tests were also included, such as urine, feces samples, and Bristol stool (BS) tests. Tests that measured the body's composition were also found, such as anthropometries and bioimpedance, to obtain parameters such as body weight (BW) and body mass index (BMI). Quality of life questionnaires included the IBD Questionnaire (IBDQ), the short IBD Questionnaire (SIBDQ), the United Kingdom version of IBDQ (IBDQ-UK), and the irritable bowel syndrome quality of life questionnaire (IBS-QOL). Table 2 schematically shows the main results found in the selected articles. Tables 3  and 4 show the scores obtained by the studies for their methodological quality according to the CONSORT and STROBE guidelines. The number of hospital admissions required in UCFR patients was 11 compared with 34 in the ND (p < 0.01). UCFR patients spent a total of 111 days in hospitals compared with 533 days for ND (p < 0.005).
Intestinal operations were performed on only one UCFR patient but on five ND.  Overall abdominal symptoms, abdominal pain, bloating, wind, and diarrhea improved in patients with CD and UC (p < 0.02 for all), but constipation did not. The median response for LFD implementation was 3/10 "easy" (SD 2.9, range 0-10, interquartile range 0.25-5 GISD, SF. and general well-being An average reduction in the total weekly score of 6.5 points was estimated for the IGED group compared with the CG (95% CI: −0.6, 13.6 points).
The estimated effect seems to have a clinically relevant effect but is not significant (p = 0.07). The daily SF significantly decreased by 11% during an IGED compared with CG.  The mean mCDAI score on entry to the trial was 171 ± 108, and after IGED, the mCDAI decreased to 97.5 ± 87 (p < 0.05). The general "well-being" rating improved from 0. 88   The reason for individuals starting the SCD was for avoidance of medication (49%), incomplete improvement with medication (28%), no improvement with medication (9%), and/or side effects or allergies to medication (19%). Overall, symptoms such as abdominal pain, limitations in activities, diarrhea, blood in the stool, and weight loss decreased over time; 4% reported clinical remission prior to the SCD, while 33% reported remission at 2 months after initiation of the SCD, and 42% reported both at 6 and 12 months Of 57 cases, 8 (4/28 IEC and 4/29 RC) relapsed during the follow-up period. Cumulative relapse rates at 1, 2, 3, 4, and 5 years were 2%, 4%, 7%, 19%, and 19%, respectively. Mean time to relapse was 7 years 3 months. There were no differences between groups. Most patients (77%) experienced some improvement. The short-and long-term PBD scores after hospitalization were higher than baseline PBD scores.  The authors did not find significant differences in either GFD or VD patient disease activities based on CDAI and MTWAI. VD patients had higher scores on the posttraumatic stress diagnostic scale and poorer mental health. A GFD was associated with lower scores in the physical and mental component survey (SF-36) and higher anxiety and depression scores.

Meta-Analysis
Only 10 clinical trials had the common quality and variables needed to be used in the meta-analysis. These 10 trials worked with a total of 13 groups. The final size of the sample was comprised of 558 observed moments for 279 individuals, all with IBD, to which a predefined diet had been prescribed. The common variables were the CDAI, FC, CRP, and ALB (Figure 3). predefined diet had been prescribed. The common variables were the CDAI, FC, CRP, and ALB ( Figure 3).
For the CDAI, which is an index of disease activity used in patients with CD [69], the effects were positive when comparing the situation at the start and at the end of treatment with a predefined diet, independently if the situation with fixed effects (less probable) or random effects (more acceptable) was considered. However, for FC, CRP, and albumin, the use of a predefined diet was not significant. As for heterogeneity, the CDAI obtained very high values, which indicates a lack of studies, and to a lesser degree, the heterogeneity is shown in CRP (80%), while for albumin (52%) and the FC (2%), the heterogeneity is not significantly high. This could indicate the high influence of some studies or the lack of them.  The influence of each study on the results of the meta-analysis is shown in Table 5, considering a model of random effects. For CDAI, the study from 2001 by Lomer et al. was the most influential; however, it is not sufficient for eliminating the high heterogeneity, and this corroborates the need for more studies or other covariables that could explain this heterogeneity. However, there are not enough studies to perform an analysis of mod- For the CDAI, which is an index of disease activity used in patients with CD [68], the effects were positive when comparing the situation at the start and at the end of treatment with a predefined diet, independently if the situation with fixed effects (less probable) or random effects (more acceptable) was considered. However, for FC, CRP, and albumin, the use of a predefined diet was not significant. As for heterogeneity, the CDAI obtained very high values, which indicates a lack of studies, and to a lesser degree, the heterogeneity is shown in CRP (80%), while for albumin (52%) and the FC (2%), the heterogeneity is not significantly high. This could indicate the high influence of some studies or the lack of them.
The influence of each study on the results of the meta-analysis is shown in Table 5, considering a model of random effects. For CDAI, the study from 2001 by Lomer et al. was the most influential; however, it is not sufficient for eliminating the high heterogeneity, and this corroborates the need for more studies or other covariables that could explain this heterogeneity. However, there are not enough studies to perform an analysis of moderators or meta-regression.
FC and ALB are not very heterogeneous, with 2.3% and 51.7%, respectively. By removing the study by Konijeti et al., 2017, the heterogeneity decreased in both cases, meaning it was the most discrepant study. Regarding CRP, the study by Chiba 2010 introduced heterogeneity at a month and a half, but at 24 months, it did not contribute. If heterogeneity is not eliminated, it can be deduced that more studies are needed. These results are reflected in the Baujat Plots ( Figure 4 Table 5.

Discussion
Our systematic review included a total of 31 studies, which compiled information from 5331 individuals with IBD and who had an intervention with different predefined diets. All the studies had a broad reach, and within the diverse effects found, CDAI, FC, CRP, and ALB were the most common, allowing us to conduct a meta-analysis to arrive at more complete conclusions.
The main premise of these types of diets was based on the reduction of some types  Table 5.

Discussion
Our systematic review included a total of 31 studies, which compiled information from 5331 individuals with IBD and who had an intervention with different predefined diets. All the studies had a broad reach, and within the diverse effects found, CDAI, FC, CRP, and ALB were the most common, allowing us to conduct a meta-analysis to arrive at more complete conclusions.
The main premise of these types of diets was based on the reduction of some types of pro-inflammatory foods and the increase of others, which are believed to promote a favorable intestinal microbiota [69]. In combination with the high prevalence of malnutrition, the importance of diets that can modify the intestinal barrier and host immunity must be increased [70,71]. In fact, although we did not observe an amelioration in terms of ALB, CRP, and CF levels, an improvement in CDAI levels was observed through interventions with predefined diets, more specifically of microparticles diet, semi-vegetarian diet, and immunoglobulin exclusion diet, in patients with CD.
The low FODMAP diet (LFD) reduces fermented oligosaccharides, disaccharides, monosaccharides, and polyols because they are poorly absorbed in the small intestine and are fermented by bacteria in the colon, triggering intestinal discomfort and gas in sensitive individuals [72][73][74]. This diet has been used mainly with patients with irritable bowel syndrome; however, it has been transferred to patients with IBD due to the similarity of functional gut symptoms such as bloating, abdominal pain, wind, and diarrhea [44,74]. As for the results obtained in our systematic review, most of the individuals improved their symptoms of the disease [44,45,59,65]. This coincides with other studies, in which an improvement was reported due to the use of an LFD for the treatment of gastrointestinal symptoms [23,75]. Furthermore, according to Pedersen [51,52,56,59]. Results of good adherence to this type of diet have also been reported [44,52,59,64], but in terms of disease activity, the results have been controversial; while for some authors no improvements were found for biomarkers or indices such as CRP, FC, HBI, or IBS-SSS, others did obtain improvements [44,51,52,56,59].
All of this, together with the concern of several authors who expressed the possibility that this type of diet may alter the microbiome by increasing the colonic pH, thereby allowing enteropathogenic colonization and causing an increase in dysbiosis [69,76,77], indicate that the use of supplementation should be considered to avoid deficiencies that could be caused by an LFD for long periods of time. Furthermore, it is of great importance that it be considered in the "induction" phase of prescription of diet modification, and if patients do not respond to the modification, the FODMAP restriction should be discontinued [76], as it can compromise the nutritional status of the patient and, to some extent, can affect intestinal inflammation [77].
The Specific Carbohydrate Diet (SCD) is based on the hypothesis that IBD patients have a dysfunction of disaccharidases, which are necessary to digest and absorb disaccharides and amylopectin. Therefore, high amounts of these compounds could cause an overgrowth of bacteria and intestinal lesions which can increase the intestinal permeability, and this is why this type of diet allows foods with carbohydrates that consist only of monosaccharides and excludes disaccharides and most polysaccharides [20]. An improvement in the symptomatology and an increase in clinical remissions are the most important results reported by Suskind et al. [56].
Both SCD and LFD have the potential to contribute to vitamin D deficiency. Therefore, their follow-up and clinical evaluation is very important due to the association of this deficiency with an increased risk of surgery and hospitalization [78][79][80][81].
The Immunoglobulin Exclusion Diet (IGED) is a dietary strategy associated with the identification of foods that cause a certain degree of intolerance, meaning an IgG-mediated reaction that acts as a delayed-type hypersensitivity response to antigen exposure, all of which result in excessive protective immune responses that could lead to increased disease activity [82][83][84]. The researchers Rajendran et al., Gunasekeera et al., and Uzunismail obtained improvements in the activity of the pathology through various tools. However, contradictory results were found for symptomatology, quality of life, and certain biochemical parameters such as CRP and ALB [39,40,48,54,67].
Several authors state that vegetarian dietary patterns are associated with a decrease in serum CRP, fibrinogen, and total leukocyte concentrations [85]. This coincides with the results obtained by Chiba et al., in which an improvement in the CRP, symptoms, and certain laboratory data could be observed [46,60]. However, it can cause an increase in posttraumatic stress and poorer mental health [66].
With respect to the Mediterranean diet (MED), characterized by the consumption of important sources of fiber (cereals, legumes, vegetables, fruits, and nuts) and with a high content of chemical compounds with antioxidant properties such as flavonoids, phytosterols, vitamins, terpenes, and polyphenols [78,86], we have obtained positive results with quality of life, HBI, FC, and cholesterol [41,58,61]. Currently, there is some controversy regarding the role of this diet in IBD, as several authors indicated that a healthy diet pattern, which includes the MED, is associated with significant reductions in inflammation-related CRP [87], and other researchers concluded that this type of diet does not have significant effects on inflammatory substances [88].
Also, there is the gluten-free diet (GFD), which eliminates the gliadin protein located in wheat, barley, rye, and other grains. This diet has been traditionally used for patients with celiac disease and more recently in people with sensitivity to non-celiac gluten [89]. However, the nutrient responsible for improvement is controversial, since these cereals have more than one possible symptom inducer such as gluten, fructans, trypsin amylase inhibitors, and lectins [90][91][92]. The results from our systematic review are controversial. On the one hand, the use of this type of diet improved the symptoms of pathology; however, it could also lead to an increase in anxiety and depression, possibly due to the difficulty of adherence [50,66]. These findings coincide with the results from some authors, who state that GFD, despite the existence of data indicating low adherence, suggests a potential benefit and great utility in the management of IBD [69,79].
Despite being the first systematic review that deals with the general effects of predefined diets on adult patients with IBD, this article is not exempt from limitations. It is possible that the CONSORT questionnaire was not the best for evaluating the Nonrandomized controlled clinical trials (NRCCT) and Uncontrolled and non-randomized clinical trial (UNRCT) reviewed; however, we tried to avoid this limitation by adjusting the items of this tool to the type of study, as no questionnaire was found that evaluated the Randomized controlled clinical trial (RCCT), the NRCCT, and the UNRCT [30,93]. Also, some studies were somewhat old, which could have reduced the score of this tool on the methodological quality due to the lack of standard criteria at the time the clinical trials were conducted. The UC and CD data were combined to perform a meta-analysis for the variables CDAI, FC, CRP, and ALB due to the low number of studies that separated these diseases to elaborate on their results and the great variability, not only of the tools used but also of the unit of measurement employed. However, these clinical entities have different clinical courses. The results derived from this work could help in clinical practice to help health professionals, through the creation of a guide oriented towards evaluating the addition of predefined diets within the set of medical therapies for an adult patient diagnosed with IBD. Both clinical trials and observational studies have been used within this systematic review, a parameter that has allowed us to have a more global view of the effect of intervention.
As future lines of research, the use of other types of predefined diets should be considered, which have been observed to show positive results in such patients and for which little evidence is found [37,38,41,42,49,50,53,60,62,63].

Conclusions
Predefined diets have been shown to have partial efficacy for the treatment of IBD and are compatible with other medical treatments. CDAI improved in patients with CD but with reasonable doubts due to the high heterogeneity of the data, while no differences were observed for ALB, FC, and CRP. LFD, IGED, MED, GFD, and vegetarian diets are the most studied and beneficial dietary interventions for these patients. However, there was a great variability in the diets and tools used to measure their interventions. In addition, the mechanisms of action of the food or nutrients responsible for the improvement are unknown. Thus, more studies that evaluate the influence of predefined diets on IBD patients are needed.