Artificial Sweeteners and Whole-Food Science: Could Mice Help Clinicians Make Diet Recommendations for IBD Patients?

Dietary therapy is emerging as a complementary strategy in some clinical guidelines to treat patients with inflammatory bowel disease (IBD),^1-4 but there is little evidence supporting what foods or drinks can be added without affecting effectiveness, and which foods should be avoided to prevent flares or intestinal inflammation. Supporting the need for diet recommendations as modulators of IBD, exclusive enteral nutrition (EEN)^1,3 has been used in IBD patients, but the benefits are not durable owing to poor palatability and compliance. Surprisingly, “sugar-free” nutrition trademarked beverages that use artificial sweeteners (ASs) and food additive combinations, possibly used to improve palatability, are listed in some guidelines as “permitted.”³

In general, clinical guidelines represent a consensus view among professionals who issue statement recommendations based on expert opinion and published literature as evidence. To support each recommendation, explicit methodologies and definitions are used to classify literature as high- or low-quality evidence. To date, experimental animal studies are not used for such evidence. High-quality evidence includes meta-analyses and randomized clinical trials (RCTs), while low-quality evidence includes studies with simpler designs, such as cross-sectional studies and case reports, which, for example, currently support antibiotic recommendations in IBD. Of note, antibiotic recommendations support selected antibiotics in specific circumstances (infection risk), despite the lack of precise understanding of the role or benefit of antibiotics in IBD,⁵ a situation that could parallel dietary recommendations (ie, low-level evidence, limited RCTs, unclear role in IBD). Overall, current clinical guidelines for IBD focus on medical and surgical approaches to induce and maintain remission. Unfortunately, no recommendations exist to specifically lower the risk of flares and recurrence of inflammation in patients who often experience worsening of symptoms despite therapy, and who often report that diet triggers symptoms.

Because the effect of diet is an emerging and complex field in IBD pathobiology and RCTs focused on ASs are currently lacking and in general are challenging to perform with diets, herein we propose to use animal feeding trials as a complementary source of data to be used as experimental-level in vivo evidence to help gastroenterologists in their recommendations for patients with IBD, and shorten the time from bench to guideline to bedside. In a field with a vast number of food ingredients that have conflicts of interest linked to industry, we also highlight the importance of reading with caution the conclusions and recommendations provided in studies and reports sponsored by the food industry.

Artificial sweeteners continue to grow as a noncaloric alternative to sugar. Since their appearance on the market with widespread human adoption to prevent weight gain, there is increasing concern based on animal studies that some of these marketed AS products have proinflammatory ingredients that may be contributing to the historic increase in IBD incidence over recent decades.⁶ At the patient level, the question is whether AS should be recommended as not to be consumed by individuals at risk for IBD. In this context, there is a need to question and discuss whether preclinical data from animal feeding trials could be used to draw recommendations on diet and ingredients.

Industry Lobbying and Unsupported Conclusions in Technical Opinions and Reports

In the case of ASs, the presence of industry representatives among coauthors in the published literature is common, and therefore there is a need to read and interpret with caution studies intended to guide clinical practice.⁷ In the United States, industry lobbying is a multibillion-dollar industry where thousands of individuals earn income from employers (industry, organizations) to promote their business models within the intended audience and influence policy making. Therefore, it is imperative to consider policy-supporting alternatives to generate scientific evidence that are less likely influenced by lobbying.

The role of reviewers and editors in the production of scientific peer-reviewed publications is of paramount importance to ensure that conclusions in manuscripts are worded as a true reflection of the data provided, free of bias. Unfortunately, at times, publications are reported with conclusions that lack scientific support within the manuscript.

Gathering human clinical evidence is complex, especially for food ingredients widely present in the food supply. In a recent publication on ASs in gastrointestinal diseases, categorized by the journal as a “guidelines and consensus statement” (and entitled by the authors as a “review of the scientific evidence and technical opinion”), one of the conclusions stated that “there is no clinical evidence of a possible inflammatory effect on the intestine caused by noncaloric sweeteners,”⁸ which may be misleading in favor of promoting the consumption of ASs. Such a conclusion, if read without careful examination of the supporting data, may be incorrectly interpreted as sufficient proof to tell IBD patients that ASs are safe to consume. However, to date, there are no clinical trials assessing the impact of ASs on IBD, and therefore such a conclusion, encompassing both naturally occurring ASs (eg, Stevia rebaudiana plant), or synthetic ASs (eg, erythritol, sucralose, acesulfame-potassium, aspartame, saccharin, cyclamate, neotame) need to be considered with caution. With no clinical trials, the lack of studies cannot be read as a lack of evidence or as evidence that there are no inflammatory effects. Why the authors, reviewers, and editors allow such unsupported conclusions is unclear, but of note, it is important to highlight that this report⁸ was “nonconditionally financed” by numerous food industry organizations (supporting the professional association and most coauthors) with active participation in AS sales worldwide. To what extent the financial conflict of interest affected the way in which the conclusions were drafted is unclear.

Focused on ASs, global leader manufacturers of consumer-packaged low-calorie sweeteners, coffee creamers, drink enhancers, and foods often use ASs in the best chemical combinations to promote better taste (“bliss point”) and product shelf-life. Such complex proprietary and trade secret combinations make the study of ASs and IBD almost impossible to generate important practical guidelines for patients, because one product may be composed of numerous ingredients widely used in the food supply. To give an example, one of several available products sold as coffee creamer lists 9 ingredients (in order of abundance: mid-oleic sunflower oil, sodium caseinate, mono-/di-glycerides, natural flavors, dipotassium phosphate, cellulose gel, cellulose gum, carrageenan, and salt), combined with 3 ASs: erythritol, sucralose, and acesulfame-potassium. Multi-ingredient proprietary combinations make it difficult to trace any one ingredient to inflammation and IBD severity.

Thus, the use of preclinical feeding trials and mouse data meta-analysis, less likely to be influenced by lobbyists, could help us understand the effect of the chemical complexity of multi-ingredient marketed products with proprietary formulations in IBD.

Not Enough Resources to Study all Dietary Ingredients in Humans

Studying every possible ingredient combination directly in humans is an insurmountable task, especially considering the factors that influence clinical study results and the limited availability of resources to sponsor the studies. One of the reasons for which no RCTs on food ingredients in human health, across diseases, exist may stem from the complexity of clinical trials and the complex role of diet composition on intestinal inflammation.⁹ The ability of clinical trials to identify true associations with disease are highly dependent on multiple factors that interfere with statistical findings. Factors include sample size and sampling strategy, genetic variability, environmental confounders, and relevance of study outcomes selected to measure the impact of diet ingredients on IBD.

We recently showed statistically how random sampling during a repeated set of similar experiments elicited irreproducible results in a large fraction of simulations, which was especially pronounced if the groups that were simulated and compared had bimodal data (2 modes/peaks, eg, high-and-low) distributions.¹⁰ In other words, comparing groups of study subjects wherein patients could have high or low disease severity, or susceptibility to diet as reflected in patient dietary surveys,^11,12 should be expected to yield variable effects on study outcomes as studies are repeated. Study findings in clinical trials are also highly dependent on the study outcome measured. For ASs, several studies have focused on examining the effect of the sweetener on the gut microbiome, glucose tolerance, lipid profiles, and obesity as outcomes, which are not directly relevant to gut inflammation in IBD.

Many food additives, including those used as fillers to provide texture for sweeteners, are widely present in the food supply and may serve as potential environmental confounders. Owing to the strong sweetening effect of ASs (>200-fold sweeter than sugar) commercial products often contain such fillers to provide volume to the product (eg, maltodextrin, which makes up 95%-99% of several marketed ASs). Unfortunately, several studies indicate that such fillers could also trigger inflammation or gut microbiome changes and thus can potentially confound clinical trials given their ubiquitous presence in the food supply and inadvertent consumption by patients. Unless diets are specifically designed and provided to study participants to control for such confounders, the use of animal models could aid in discovering how ASs alone or in commercial combinations influence IBD severity.

IBD Genetics and Animal Models as an Acceptable New Level of Evidence

Chronic exposure of inflammatory mediators are well known risk factors for chronic conditions, but susceptibility in humans is highly dependent on genetics, which remains largely uncharacterized. Because the prevalence of predisposition to certain diseases tends to be relatively rare (ie, 79-90/100,000 globally and 439-491/100,000 in US suffer from or are prone to IBD),⁶ such susceptibility may not be represented in clinical studies owing to random sampling heterogeneity, which is not necessarily compensated for by large sample sizes.¹⁰ The study of factors that could modulate chronic diseases which frequently alternate between remission and flare-ups in complex clinical phenotypes (eg, Crohn’s disease ileitis vs colitis vs ulcerative colitis) is further complicated with the uncertain distribution of hundreds of genetic loci (>200) that are epidemiologically associated with less than 17% of IBD susceptibility.¹³

The characterization of susceptibility to diets as a function of genetics would be more cost-effective and reproducible in animal models. Unfortunately, most studies on ASs have been conducted with “healthy” animals, including mouse lines that are not prone to spontaneous IBD (eg, C57BL/6, Swiss Webster). Our group has shown that Splenda (sucralose + maltodextrin) supplemented in drinking water for 6 weeks had significant outcomes on ileitis in mice prone to IBD (SAMP1/YitFc) with increased proinflammatory myeloperoxidase activity, Proteobacteria gut microbiome dysbiosis, and gut bacteria penetration in gut wall.¹⁴ In contrast, such inflammatory changes were not observed in healthy control mice (AKR/J) despite changes being observed in the gut microbiome. That study¹⁴ and others¹⁵ (Table 1) show that changes in the gut microbiota induced by ASs are not necessarily correlated with IBD inflammation unless the consumer has genetic susceptibility to IBD.

Table 1.

Example of Animal Studies Assessing AS on Intestinal Homeostasis and Inflammation^a

Study	Sweetener and Dosage	Rodent and Diet	Effect of AS
Li 201619	AS: sucralose or saccharin. C: none specified. D: 0.3, 1.0, and 1.5 mg/mL in water for 4-6 weeks. Single dose of 1.0 mg/mL.	Rats, unspecified. Standard mouse diet.	Increased digestive proteases fecal chymotrypsin and trypsin (dose-dependent for sucralose) with decrease in fecal β-glucuronidase.
Bian 201720	AS: sucralose pure. C: tap water. D: 0.1 mg/mL (equivalent to ADI of 5mg/kg/day in humans) for 6 months.	Male C57BL/6 mice. Standard mouse diet.	Host microbiota and related metabolites alteration, bile acids. Increased genes on bacterial proinflammatory mediators (LPS, flagella protein synthesis, fimbriae, shiga toxin); disrupted quorum sensing signaling; liver inflammation. Altered host metabolites, ie, tryptophan, quinolinic and kynurenic acid. Tyrosine p-hydroxyphenyl acetic and cinnamic acids (ROS production inhibitors).
Bian 201721	AS: saccharin. C: tap water. D: 0.3 mg/mL in water for 6 months.	Male C57BL/6 mice. Standard mouse diet.	Increased inflammation factors; iNOS enzyme, TNFα in liver. Alteration of gut microbiota with enriched orthologs of pathogen associated molecular patterns such as LPS, flagellar assembly, and bacterial toxins. Increased proinflammatory metabolites
Rodriguez-Palacios 201814	AS: sucralose (Splenda). C: plain water. D: 1.08–3.5 mg/mL in water for 6 weeks.	SAMP1/YitFc, C57BL/6, AKR mice. Irradiated standard diet.	Intestinal dysbiosis (Proteobacteria expansion), and increased ileal MPO activity. Genetic susceptibility: effects more pronounced in IBD-prone SAMP1/YitFc mice.
Wang 201922	AS: sucralose. C: none specified. D: 1.5 mg/mL in water for 6 weeks before TNBS colitis induction. Design not well described.	Sprague-Dawley rats. Standard rodent diet.	Sucralose increased gut damage, permeability (serum d-lactic acid) and inflammation (MPO, TNFα, and IL1β) with/without TNBS-induced colitis, and increased digestive proteases fecal chymotrypsin and trypsin; decreased β-glucuronidase in feces.
Martinez-Carrillo 201923	AS: sucrose and Splenda (sucralose) or Svetia (stevia). C: plain water. D: 41.6 mg/mL and 4.1 mg/mL of commercial AS formulations in water (5 h/day) for 6–12 weeks.	CD1 mice. Standard rodent diet.	Intestinal dysbiosis and increased percentage of lymphocytes and IL6 and IL17 in Peyer’s patches and lamina propria after consumption of Splenda and stevia. Splenda and stevia increased leptin and C-peptide whereas TNFα decreased with Splenda but not stevia.
Guzman-Cruz 201915	AS: sucralose (Splenda) and sucrose or stevia (Svetia). C: plain water. D: 41.6 mg/mL and 4.1 mg/mL of market AS formulations in water (5 h/day) for 6 weeks.	Male CD1 and Balb/c mice.	CD1 mice had higher percentage of lymphocytes in Peyer’s patches after consuming Splenda and stevia compared to Balb/c mice. Genetic susceptibility: In Balb/c, sucrose increased the percentage of lymphocytes in Peyer’s patches whereas sucrose reduced this percentage in CD1 mice.
Farid 202024	AS: sucralose (Sweetal), stevia (SweetLeaf), or sucrose. C: plain water. D: 5.2 mg/mL, 4.2 mg/mL, 40.5 mg/mL in water (5 h/day) for 8-16 weeks.	Male/female BALB/c albino mice. Standard rodent diet.	Elevation in HbA1c and different immunoglobulins (IgG, IgE, IgA) and proinflammatory cytokines (IL6, IL8), with reduction in antiinflammatory cytokine IL10 in mice fed sucralose or stevia; alteration in kidney and liver function. Sex differences noted.
Li 202025	AS: sucralose. C: not specified. D: 1.5 mg/mL in water before (6 weeks) and during AOM and DSS (2 cycles) treatment.	C57BL/6 mice. Standard rodent diet.	Intestinal dysbiosis and increased tumorigenesis, and worsening of DSS severity, including gut barrier integrity, with impaired inactivation of digestive and decrease in fecal β-glucuronidase.
Sanchez-Tapia 202026	AS: sucralose, steviol glycosides (SG), or SG + sucrose. C: sucrose, glucose, fructose, honey, or brown sugar. D: 1.5%, 2.5%, or 10% in water for 4 months	Male Winstar rats. Standard rodent diet or HFD.	Reduced gut microbiota diversity. Type of sweetener and an HFD changed metabolic profiles and microbiota. Sucralose decreased occludin and increased proinflammatory cytokines and bacterial genes involved in synthesis of LPS and SCFAs, glucose intolerance, fatty acid oxidation; HFD exacerbated effect.

ADI, acceptable daily intake; AS, artificial sweetener; C, control; D, dosage; HbA_1c, glycated hemoglobin; HFD, high-fat diet; IL, interleukin; iNOS, inducible nitric oxide synthase; LPS, lipopolysaccharide; MPO, myeloperoxidase; ROS, reactive oxygen species; SCFA, short-chain fatty acid; TNBS, 2,4,6-trinitrobenzenesulfonic acid; TNF, tumor necrosis factor.

^aCurrently, there are 6 noncaloric artificial sweeteners on the market: aspartame, saccharin, sucralose, acesulfame-potassium, cyclamate, and neotame. In terms of the gut microbiota, sucralose, saccharin, stevia, and polyols appear to shift the gut microbiota in mice.

Because studying the functional effect of individual ingredients directly in humans is virtually impossible in this context, with current tools, the use of other animals represents the closest viable alternative to study the causal association between food ingredients and IBD, especially as mice share a great percentage of genome features and immunologic pathways that are similar, annotated, and functionally conserved and cross-referenced with humans.¹⁶ Rodent studies also take advantage of lower cost, controlled environment, fewer ethical implications, shorter life span, and easier replication. Despite these benefits, a few challenges associated with sample size and other factors that we and others have examined, need to be addressed.¹⁷

In short, there is a need to improve the quality of mouse studies to correct for technical confounders that interfere with laboratory studies¹⁸ by 1) standardizing the reporting of diet, 2) preventing cage-cage microbiome heterogeneity before experiments start, 3) preventing periodic cyclical bias and variations in gut microbiome due to accumulation of excrements/humidity in cages, 4) controlling for seasonal variability, 5) controlling for cage cluster effects, and 6) statistically use study power and control for cage clustering to maximize reproducibility. To standardize animal feeding trials, we suggest using IBD-relevant outcomes (eg, histology/biochemical inflammation; gut wall bacterial penetration) to simulate clinical trials, using larger sample sizes (eg, 20-50 mice instead of <6),¹⁸ and focusing on examining the effect of diet on inflammation as primary outcomes. Such a primary outcome should be used to estimate sample sizes, study power, and magnitude of effects. Single ingredient–based mechanisms in IBD or other organs should be examined as secondary outcomes. The creation of a registry for mouse feeding trials could also help build a body of evidence to connect diet and genetic susceptibility via mouse data metanalyses.

We suggest that when there is reproducible preclinical experimental evidence showing potential adverse consequences to gut health, IBD patients should be recommended not to consume a given diet instead of to use with caution until appropriate RCTs are performed. “Use with caution” is noninformative and nonguiding to help patients because there are no objective means to determine if any given diet/ingredient (eg, AS) is good or bad for IBD. A graphical overview of the current scenario and a potential future strategy to design and use animal feeding trials to support recommendations for patients as a form of experimental in vivo level of evidence for IBD is presented in Figure 1. For illustrative purposes, Table 1 summarizes the effect of ASs (namely sucralose) on gut health in mice with and without susceptibility to IBD. Owing to the comparable direction of effects summarized in the table, most of which are adverse for gut homeostasis and inflammation vs control diets, this animal data could be used to support a recommendation that IBD patients should not use sucralose, in contrast to IBD-free individuals, where adverse IBD effects are not expected.

Figure 1.

Single dietary ingredient vs whole-food science: graphical overview of how mouse data meta-analyses could help clinicians make diet recommendations for IBD patients. Studying every possible combination directly in humans, if we rely only on the use of single-ingredient science approach, would likely be an insurmountable task, especially considering the factors that influence clinical study results. Standardizing the quality and reporting of animal studies will ultimately enable meta-analysis of animal data.¹⁸

In conclusion, the science of food and health is still a growing field with an almost infinite number of combinations and disease outcomes to be studied. Individual-ingredient science cannot be translated to food categories, because all foods are composed of many ingredients or chemicals. Despite several studies, thus far, none of the mouse data for single ingredients or ASs has been translated to practical recommendations to benefit IBD patients. Animal data should therefore not be discounted by clinicians. The examination of diets as a whole-food science on in vivo model systems could enable the rapid cost-effective generation of diet recommendations for IBD. In addition to highlighting that IBD therapy and surgery must rely on RCT evidence to make recommendations, we suggest that for prevention of diet-driven symptoms reported by surveyed patients, it is important to consider studies derived from animals to raise awareness and create prevention guidelines that could be updated regularly to help individuals with IBD. Ultimately, if we make the leap from animal modeling to patients it remains important to follow up with outcome-based clinical studies in humans. Meanwhile, mouse data and meta-analyses could be the next generation of experimental-level in vivo evidence to help create such guidelines where there is a gap in knowledge^1-4 to help document and expand current recommendations.