Effects of the Non-Nutritive Sweeteners on Glucose Metabolism and Appetite Regulating Hormones: Systematic Review of Observational Prospective Studies and Clinical Trials

Background The effects of non-nutritive sweeteners (NNS) on glucose metabolism and appetite regulating hormones are not clear. There is an ongoing debate concerning NNS use and deleterious changes in metabolism. Objectives The aim of this review is to analyze the scientific available evidence regarding the effects of NNS on glucose metabolism and appetite regulating hormones. Data Sources and Study Eligibility Criteria We identified human observational studies evaluating the relation between NNS consumption and obesity, diabetes, and metabolic syndrome, in addition to clinical trials evaluating the effects of NNS in glucose metabolism and appetite regulating hormones. Results Fourteen observational studies evaluating the association between NNS consumption and the development of metabolic diseases and twenty-eight clinical trials studying the effects of NNS on metabolism were included. Finally, two meta-analyses evaluating the association between the consumption of NNS-containing beverages and the development of type 2 diabetes were identified. Conclusions Some observational studies suggest an association between NNS consumption and development of metabolic diseases; however, adiposity is a confounder frequently found in observational studies. The effects of the NNS on glucose metabolism are not clear. The results of the identified clinical trials are contradictory and are not comparable because of the major existing differences between them. Studies evaluating specific NNS, with an adequate sample size, including a homogeneous study group, identifying significant comorbidities, with an appropriate control group, with an appropriate exposure time, and considering adjustment for confounder variables such as adiposity are needed.


Introduction
The prevalence of obesity has more than doubled since 1980; in parallel in 2014, the estimated number of patients with diabetes in the world was 385 million with a projection to increase to 592 million by 2035. One of the contributing factors attributed to the increase in obesity, type 2 diabetes and other metabolic diseases is the consumption of a high sugar/high fat diet [1]. To avoid the negative health conditions associated with the excessive sugar intake, there has been an upsurge in the consumption of nonnutritive sweeteners (NNS) as an alternative [2]. At this time six NNS, sucralose, aspartame, saccharin, acesulfame-K, neotame, and advantame, are approved to be used as sweeteners in food, and two naturally derived NNS, steviol glycosides and Luo han guo extract, are generally recognized as safe and endorsed for use in food by the US Food Drug Administration (FDA) and the European Food Safety Authority (EFSA) [3,4]. Nowadays, they are globally used and they are found in several products.
Recently, the EFSA conducted a re-evaluation of aspartame safety, and concluded that aspartame and its breakdown products are safe for the general population (including infants, children and pregnant women) [4]. Before the FDA approved NNS consumption, a series of toxicological and clinical studies in a number of species, including humans, were conducted to demonstrate that they are generally safe and well-tolerated [5]. There is an ongoing debate over whether NNS use may be associated to deleterious metabolic changes in humans [6]. This article aims to collect the information regarding the effects of NNS consumption on metabolic diseases, based on a systematic review of the scientific literature.

Study Search and Selection
We identified human studies evaluating the effects of NNS consumption in metabolic conditions through systematic searches and hand searches on April 8, 2015 (updated on March 25, 2016) in three electronic databases: PubMed, The Cochrane Library, and Trip Database. We conducted the search for observational studies to answer the following research question: Is there a relation between NNS consumption and the development of metabolic chronic diseases in adults? For clinical trials, we directed the search to answer the next research question: Is there an effect of NNS on glucose metabolism and appetite regulating hormones compared to water or other sweeteners in adults? The terms used in the systematic search were those related to NNS and artificially sweetened beverages including the next Medical Subject Headings (MeSH) terms: artificial sweeteners / non-nutritive sweeteners / carbonated beverages / sucralose / aspartame / stevia / saccharin / acesulfame potassium / diet soda / diabetes mellitus / obesity / metabolic syndrome. To complement the search, we also performed a hand-searching strategy through certain journals and references in other articles. Time and language of publication were not restricted. Inclusion criteria consisted in original studies of prospective design conducted in adult humans. For cohort studies, we considered those that evaluate the association between NNS consumption and the development of diabetes, metabolic syndrome or obesity, with a follow up of at least three years. For clinical trials we included those that evaluate the effects of any NNS on outcomes related to glucose metabolism and appetite regulating hormones (S1 File). One researcher (AR) screened the articles titles and abstracts to remove those that easily were detected to be not related to the objective of this review, and three researchers (AR, PA, and GB) read the articles that could be eligible in the systematic review and select those that finally are included. Articles evaluating the effects of NNS in other conditions or evaluating other outcomes not related were excluded.

Literature search
We identify 1569 studies through database searching; in addition, 376 were found by the hand searching strategy. After duplicates removal and initial screening, 72 studies were reviewed. Finally, 44 studies were included after the exclusion of 28 that did not fulfill the inclusion criteria.

Observational studies
We included fourteen observational studies evaluating the association between NNS consumption and the development of metabolic diseases including type 2 diabetes, obesity, and metabolic syndrome. All of these studies have considered NNS consumption in beverages and most of them in soft drinks.
Summarizing the results, the majority of these studies have found significant associations between the ingestion of NNS and the development of metabolic diseases. Among these studies there are two reports derived from the Nurses' Health Study (NHS I and II) that included more than 70,000 and 90,000 women, with an average follow-up of 24 and 8 years for the first and the second studies, respectively. The first of these studies found a significant association between caffeinated artificially sweetened beverages consumption and development of type 2 diabetes (RR 1.35, 95% CI 1.24-1.47). However, this association was lost after the adjustment for body mass index (BMI) and energy intake (RR 1.01, 95% CI 0.93-1.10) [7]. In the NHS II no association was found [8].
Another large cohort study that evaluated the effect of artificially sweetened beverages consumption and the development of type 2 diabetes is the Health Professionals Follow-Up Study. This included approximately 40,000 male health professionals followed over 20 years. This study found a significant association between NNS consumption and type 2 diabetes development, even after multivariable adjustment (HR 1.40, 95% CI 1.26-1.56). However, this association was lost after the adjustment for BMI (HR 1.09, 95% CI 0.98-1.21) [9].
The European Prospective Investigation into Cancer and Nutrition (EPIC) Study, performed in eight European countries, included 340,234 men and women. This study reported a significant association between artificially sweetened soft drinks ingestion and type 2 diabetes development (HR 1.93, 95% CI 1.47-2.54). This association was attenuated after multivariable adjustment (HR 1.88, 95% CI 1.44-2.45), and lost statistical significance after further adjustment for BMI and energy intake (HR 1.13, 95% CI 0.85-1.52) [10]. Table 1 shows a summary of the results of the included cohort studies. On Table 2 the crude and adjusted risks reported in these studies are contrasted.

Clinical trials
Twenty-eight clinical trials studying different effects of NNS were identified. Of these studies, 10 found significant effects on some or all the studied variables. All of these studies have analyzed glucose and most of them have measured insulin concentrations, 11 quantified GLP-1 concentrations. However, only one study has measured insulin sensitivity and pancreatic response, and another single study has evaluated the changes in the intestinal microbiome. The majority of the clinical trials have evaluated the effects of aspartame (14 trials), followed by sucralose (11 studies), and saccharin, acesulfame-K, and stevia (5 studies for saccharin, 5 for acesulfame-K, and 4 for stevia). Most of these studies have performed an acute single exposure to the NNS (n = 20) and the remaining (n = 8) have evaluated a longer exposure that varies between seven days to 18 weeks. Thirteen studies included individuals with diabetes.
The studies by Pepino [20] and Suez [21] demonstrate a deleterious effect increasing glucose concentrations after an acute and a 7-day exposure to sucralose and saccharin, respectively. Pepino, also reported a decrease in insulin sensitivity along with increased insulin and C-peptide concentrations. Remarkably, this study included subjects with a high degree of obesity (average BMI 42 kg/m2). In the study of Suez after a seven-day period of saccharin ingestion, in four of seven subjects glucose concentrations showed a significant increment. Subsequently, a feces transplant from some of the individuals with the glucose increase after saccharin exposure to mice was performed. After the transplant, glucose concentrations also increased in these mice, suggesting that NNS consumption modify intestinal microbiome in detriment of glucose tolerance. The microbiome showed a significant imbalance with an increase in the Bacteroides genus and Clostridiales order [21]. GLP-1 concentrations, measured in eleven studies, have been shown to be decreased in one report after aspartame ingestion [22] and increased in two studies after sucralose + acesulfame-K and sucralose exposure [23,24]. Concentrations of appetite-regulating hormones, including cholecystokinin, ghrelin, and peptide YY, have only been studied in three studies. In none of them changes in the concentrations of these variables were found. In addition, no change in the subjective appetite ratings or on the quantity of food consumed after NNS exposure has been found. On Table 3 the description and results of these studies are shown. As a reference, one 12 oz diet-coke contains approximately 140 mg of aspartame and acesulfame K mix, one 12 oz diet-Dr. Pepper can contains approximately 65 and 22 mg of sucralose and acesulfame, respectively, and one 12-oz Coca-Cola Life can contains 27 mg of stevia. Some of the NNS available as individual packets include Sweet and Low, containing 34 mg of saccharin, and Splenda containing 12 mg of sucralose. On Table 4 a summary of the studies indicating the methodology used, studied variables, and the NNS evaluated are presented.

Meta-analysis
Two meta-analyses have been published to evaluate the association between the consumption of NNS-containing beverages and the development of type 2 diabetes to clarify if this relation is clearly linked to the consumption of these products or related to other lifestyle factors. Both meta-analyses evaluated the association between NNS consumption, without specifying of stratifying for the specific NNS ingested. While both studies excluded cohorts including individuals with a known diagnosis of diabetes, the article by Grenwood only included four studies. This may be due to the selection criteria that specify that only studies including individuals "from a generally healthy population" were considered [48]. In contrast, the study by Imamura evaluated ten studies estimating the risk of type 2 diabetes associated to consumption of NNScontaining beverages [49]. None of the studies disclosed significant competing interests.
In the first meta-analysis that included 4 observational prospective studies, the pooled estimated relative risk (RR) was 1.13 (95% CI: 1.02-1.25; P = 0.02) for the consumption of 330 ml per day of artificially sweetened beverages and the development of type 2 diabetes. There was   • 24 subjects aged 20-50 years with obesity (BMI 28-36 kg/m 2 ) • Individuals with diabetes or pregnancy were excluded • Crossover study • 4 visits: 500 ml sucrose-sweetened regular soda, 500 ml semi-skimmed milk, 500 ml aspartame-sweetened diet soda or 500 ml bottled still water. • Ad libitum food consumption from a buffet after 4 hours • VAS to measure appetite • Subjects aged 12-25 years divided in 3 groups: 9 with T1D, 10 with T2D, and 25 healthy control participants • All T2D were overweight or obese • Crossover study • 3-h OGTT with 75 g • 2 visits: at minute -10 subjects drank 240 ml of diet soda with sucralose and acesulfame-K or 240 ml of carbonated water • GLP-1 AUC 43% higher with the ingestion of diet soda in T1D subjects (p = 0.02) • GLP-1 AUC 34% higher with the ingestion of diet soda in healthy subjects (p = 0.029) • No differences on glucose, C-peptide, GIP, and PYY (Continued) Non-Nutritive Sweeteners, Glucose Metabolism, and Appetite Hormones high heterogeneity between studies, and the positive association was less consistent for this type of beverages compared to the sugar-sweetened drinks [48].

Discussion
The aim of this systematic review is to evaluate the scientific available evidence regarding the association between NNS consumption and metabolic diseases as well as the effects of NNS on glucose metabolism and appetite regulating hormones. The results indicate that the association between NNS intake and the development of metabolic diseases, mainly type 2 diabetes, is not clear. A common identified confounding factor in the observational prospective studies is adiposity. In addition, it is unknown if the NNS are associated with deleterious effects on glucose metabolism or appetite regulation. Based on the available evidence, an effect of NNS on glucose metabolism cannot be established. The study of appetite and its regulation is complex, the evidence presented concerning this issue is scarce and an effect of NNS in appetite cannot be demonstrated either. The studies found are varied regarding the NNS studied; therefore, a class effect cannot be determined and no solid conclusions regarding a specific NNS can be stated. A possible explanation for the associations found in some of the observational studies among NNS consumption and the development of metabolic diseases might be that these cohorts included participants prone to develop these outcomes, for example with family members with diabetes or with a predisposition for weight gain, that are likely to consume these Table 4. Summary of the studied variables, non-nutritive sweetener used, study methodology and findings of the clinical trials evaluated in Table 3.
An additional limitation of observational studies is that the majority has evaluated the consumption of NNS containing-beverages and the non-consuming population may actually have consumed these substances from other non-acknowledged products. Finally, the evidence level of observational studies cannot establish causality.
Most of the clinical trials included have small sample sizes and the majority does not provide a justification for these calculations. Many of the clinical trials are crossover studies and a main limitation of this design is the residual effect between treatments. In most cases there is no information regarding the washout period. Another variable that needs to be considered is the amount of NNS used and the exposition type, for example acute or long-term exposition. Moreover, there is no uniformity in the exposition time between studies evaluating a long-term exposure. Finally, a number of confounding variables are not mentioned or adjusted in these trials, including BMI, previous NNS intake, and presence of metabolic alterations such as glucose intolerance or diabetes, among others. These drawbacks may confuse the results presented.
We can conclude that some clinical trials have found effects of NNS on glucose metabolism. However the results are contradictory and there is no possible comparison between the trials due to the heterogeneity in the population included, NNS studied, placebo use, exposure time, outcomes evaluated, among many other. For example, after sucralose consumption, one study reported higher concentrations of glucose, however, another study report lower concentrations and nine studies did not observed changes in glucose. In addition, two studies found that sucralose increase GLP-1 levels compared to water, an effect that other six studies could not confirm. One study found that sucralose decreases insulin sensitivity and insulin clearance in morbid obese population, nevertheless, this is the only one trial that has evaluated these outcomes.
The consumption of aspartame showed lower concentrations of glucose in two of fourteen studies, one compared to water and the other one to glucose. One study observed lower concentrations of insulin after aspartame vs. sucrose and another study found higher concentrations of insulin after aspartame vs. water. Finally, one trial reported that aspartame decreases GLP-1 concentrations compared to placebo.
For stevia, one trial observed lower glucose and insulin concentrations compared to sucrose, and another study found lower glucose concentrations and an increment in the insulinogenic index compared to placebo.
One trial reflected an important impact of saccharin consumption for seven days promoting glucose intolerance in four of seven subjects studied; this trial suggest that this effect is caused by altering the gut microbiome performing a fecal transplantation to mice showing a similar increase in glucose levels. In contrast, one study showed lower glucose and insulin concentrations after saccharin ingestion compared to water or sucrose, respectively.
The findings of the two meta-analyses should be interpreted cautiously. In the first report few studies were included, without considering other variables that may be involved in the development of diabetes, and in the second the association between NNS-containing beverages and development of diabetes was lost after the adjustment for body mass index, indicating that adiposity may be influencing the findings.
Randomized clinical trials testing each of the NNS, including a homogeneous group of participants, without metabolic conditions that may confound the results, including an adequate sample size, with an appropriate control group, during an appropriate exposure time, and considering adjustment or control for significant variables such as adiposity are needed. In addition, the mechanisms involved in the glucose metabolism changes after a long-term exposition to NNS should explored in human studies.
Based on the scientific evidence presented, the consumption of NNS is not encouraged, but they could be considered a useful tool in the nutritional treatment of certain metabolic diseases as sugar substitutes as long as the quantity consumed is within the acceptable daily intake (ADI) and without compensation by ingesting other energy-rich foods. Lastly, health professionals should not promote the consumption of sweet tasting foods regardless its source.

Author Contributions
Conceived and designed the experiments: AR PAV.
Performed the experiments: AR.