Nonnutritive sweeteners (NNSs) are zero- or low-calorie alternatives to nutritive sweeteners, such as table sugars. A systematic review and meta-analysis of randomized controlled trials was conducted to quantitatively synthesize existing scientific evidence on the glycemic impact of NNSs.
PubMed and Web of Science databases were searched. Two authors screened the titles and abstracts of candidate publications. The third author was consulted to resolve discrepancies. Twenty-nine randomized controlled trials, with a total of 741 participants, were included and their quality assessed. NNSs under examination included aspartame, saccharin, steviosides, and sucralose. The review followed the PRISMA guidelines.
Meta-analysis was performed to estimate and track the trajectory of blood glucose concentrations over time after NNS consumption, and to test differential effects by type of NNS and participants’ age, weight, and disease status. In comparison with the baseline, NNS consumption was not found to increase blood glucose level, and its concentration gradually declined over the course of observation following NNS consumption. The glycemic impact of NNS consumption did not differ by type of NNS but to some extent varied by participants’ age, body weight, and diabetic status.
NNS consumption was not found to elevate blood glucose level. Future studies are warranted to assess the health implications of frequent and chronic NNS consumption and elucidate the underlying biological mechanisms.
Obesity is a leading cause of morbidity and premature mortality in the U.S. and worldwide . From 1976–1980 to 2013–2014, the prevalence of obesity more than doubled among the American adult population . It is estimated that the medical expenditure attributable to overweight and obesity will reach $861–957 billion U.S. dollars by 2030, accounting for 16–18% of the nation’s total health care costs . By 2025, the global obesity prevalence is expected to reach 18% in men and exceed 21% in women, and severe obesity will exceed 6% in men and 9% in women .
A long line of scientific research documented the risk of sugar consumption (in forms of free sugars or sugar-sweetened beverages) for childhood and adult obesity . Sugar consumption may increase energy intake to an extent that exceeds energy output and distorts energy balance . Energy in liquid form, such as sugar-sweetened beverages, can be less satiating than when derived from solid foods, resulting in overconsumption . Solid foods high in sugar are often energy dense, and frequent consumption of such foods may lead to weight gain . Restricting caloric intake from sugar has become an important public health strategy for weight management and obesity prevention. The 2015–2020 Dietary Guidelines for Americans recommends <10% of daily calories from added sugars . The World Health Organization’s Guideline: Sugars Intake for Adults and Children recommends reduced consumption of free sugars throughout the life course and suggests a further reduction of sugar consumption to below 5% of daily calories .
Nonnutritive sweeteners (NNSs), also called artificial sweeteners, are zero- or low-calorie alternatives to nutritive sweeteners (e.g., table sugar). NNSs provide a sweet taste with an addition of few calories, making it a popular substitute to sugars. From 1999–2000 to 2009–2012, NNS consumption in the U.S. increased by 200% in children and 54% in adults . In 2009–2012, 25% of children and 41% of adults reported consuming NNSs on a regular basis . Currently eight NNSs are permitted for use in food in the United States, including saccharin, aspartame, acesulfame-k, sucralose, neotame, advantame, steviol glycosides, and Luo Han Guo fruit extracts . These NNSs are high-intensity sweeteners 30 to 20,000 times sweeter than sucrose, which allows for less NNS use in product formulations when replacing sugar . Although the Food and Drug Administration and most published studies endorse the safety of these NNSs, conclusive evidence has yet been drawn regarding their use on a regular basis . The clinical and epidemiologic data at the current stage remain insufficient to make definitive conclusions regarding the benefits of substituting sugars with NNSs on energy balance and weight management . Nevertheless, moderate consumption of NNSs is thought to be beneficial as a dietary aid for people with diabetes or on a weight loss routine .
Unlike sugar consumption that raises glycemia—glucose level in the blood—through metabolism, it is widely believed that NNSs may not have a substantial impact on glycemia . However, different NNSs possess diverse chemical structures and distinct post-ingestive behaviors, resulting in some NNSs undergoing metabolism . Recent research indicates differential effects of sucralose, a type of NNS, on glucose metabolism between normal weight people and people with obesity . A recent review examined the effects of NNSs on glucose metabolism and appetite-regulating hormones . However, it mainly focused on the evidence linking NNS consumption to the risks for chronic diseases including obesity, diabetes, and metabolic syndrome rather than the short-term response of blood glucose level to NNS consumption . The purpose of the current study was to systematically review and quantitatively synthesize existing scientific evidence from randomized controlled trials (RCTs) on the glycemic impact of NNSs. Meta-analysis was performed to estimate and track the trajectory of glucose concentrations over time after NNS intake, and test differential effects by type of NNS and participants’ age, weight, and diabetic status.
Materials and methods
Study selection criteria
Studies that met all of the following criteria were included in the review—Study design: RCT; Study subject: human; Intervention: oral NNS consumption after overnight fasting; Outcome: change in blood glucose level in response to NNS consumption; Article type: peer-reviewed publication; and Language: English.
Studies that meet any of the following criteria were excluded: (1) studies that administered NNS consumption in combination with other caloric foods or beverages, (2) studies not explicitly reporting a fasting protocol, (3) studies reporting no results from blood glucose test, or (4) presence of other intervention components that may affect blood glucose (e.g., exercise).
A keyword search was performed in two bibliographic databases—the PubMed and the Web of Science. We conducted the searches from the time of database inception to 23 February 2018. The search algorithm included all possible combinations of keywords (with wildcards) from the following two groups: (1) “non-nutritive sweetener”, “rebaudioside B”, “nonnutritive sweetener”, “non nutritive sweetener”, “artificial sweetener”, “natural sweetener”, “low calorie sweetener”, “low-calorie sweetener”, “zero calorie sweetener”, “zero-calorie sweetener”, “stevia”, “saccharin”, “aspartame”, “trichlorosucrose”, “sucralose”, “acetosulfame”, “acesulfame*”, “neotame”, and “rebaudioside A”; and (2) “glucose tolerance*”, “glycemic load”, “glycaemic load”, “glucose load”, “blood glucose”, “oral glucose”, “hyperglycemia”, “hypoglycemia”, “hypoglycaemia”, and “intravenous glucose”. Titles and abstracts of the articles identified through the keyword search were screened against the study selection criteria. Potentially relevant articles were retrieved for evaluation of the full text. Cohen’s kappa (0.88) was used to evaluate interrater agreement. Two authors of this review, ADN and MJH, jointly determined the inclusion and exclusion of all articles retrieved in full text, and discrepancies were resolved through consultation with the third author RA.
A standardized data extraction form was used to collect the following methodological and outcome variables from each included study: author(s), publication year, NNS type, blood glucose levels, sample size, and participants’ characteristics (i.e., age, gender, body weight status [body mass index (BMI)], and diabetic status).
Meta-analysis was performed to estimate the glycemic impact of NNS consumption. All reported blood glucose values were converted to a standard unit of mmol/L. Blood glucose measurements were allocated to 30-min intervals starting from the time of NNS consumption (i.e., baseline) to 210 min after consumption. The outcome variable was the change in blood glucose level relative to the baseline. If a study reported multiple glucose values in a single 30-min interval, those values were averaged with pooled standard error calculated. Study heterogeneity was assessed using the I2 index. The level of heterogeneity represented by the I2 index was interpreted as modest (I2 ≤ 25%), moderate (25% < I2 ≤ 50%), substantial (50% < I2 ≤ 75%), or considerable (I2 > 75%). A fixed-effect model would be estimated when modest to moderate heterogeneity was present, and a random-effect model would be estimated when substantial to considerable heterogeneity was present. Publication bias was assessed by the Egger’s test. Meta-regressions were conducted to assess the potentially differential glycemic impacts by type of NNS and participants’ age, weight, and diabetic status. All statistical analyses were conducted using Stata 14.2 SE version (College Station, TX: StataCorp LLC). All analyses used two-sided tests and p-values <0.05 were considered statistically significant.
Study quality assessment
Adapted from the National Institutes of Health’s Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies, quality of the included studies was assessed using seven criteria: (1) Was the research question clearly stated? (2) Were the inclusion and exclusion criteria clearly stated? (3) Were study participants’ body weight status (height/weight and/or BMI) objectively measured and clearly reported? (4) Was a sample size justification via power analysis provided? (5) Was the dropout rate from the study 20% or lower? (6) Was the population referenced in the conclusion appropriate? (7) Were there 20 or more study participants that received NNS without additional caloric intake? The two authors of this review, ADN and MJH, independently scored each study based on these seven criteria, with disagreement resolved through discussion. Scores for each criterion range from 0 to 2, depending on whether the criterion was unmentioned or unmet (0), partially met (1), or completely met (2). The total study scores range between 0 and 14. Study quality assessment helped measure the strength of scientific evidence but was not used to determine the inclusion of studies.
Figure 1 shows the study selection flowchart. Among 460 total unduplicated articles identified through the keyword and reference search, 205 were excluded by title and abstract screening. The full texts of the remaining 255 articles were reviewed, and 226 were excluded for not meeting the study selection criteria. Main reasons for exclusion included studies not explicitly reporting a fasting protocol, NNS consumption in combination with other caloric foods or beverages, no results from blood glucose test, or presence of other intervention components that may affect blood glucose (e.g., exercise). The remaining 29 articles [20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48] were included in the review.
Basic characteristics of the selected studies
Table 1 reports the basic characteristics of the 29 articles included in the review. Eleven studies evaluated aspartame, 12 evaluated saccharin, 3 evaluated stevia or steviosides, and 5 evaluated sucralose. Among a total of 741 study participants, 406 were normal weight (18.5 kg/m² ≤ BMI < 25 kg/m²), 71 were overweight (25 kg/m² ≤ BMI < 30 kg/m²), 20 were obese (BMI ≥ 30 kg/m²), and the remaining 244 did not report their BMI. Regarding NNS consumption, 280 participants consumed aspartame, 214 consumed saccharin, 105 consumed stevia, and 142 consumed sucralose. Regarding health status, 526 participants reported no chronic condition, 69 reported type 2 diabetes, and 146 did not report their disease status. Age across all studies ranged from 9 to 69 years old, with an average of 38 years old.
Estimated glycemic impact of NNS consumption
Figure 2 shows the trajectory for the glycemic impact of NNS consumption estimated in the meta-analysis. In comparison with the baseline, NNS consumption was not found to increase blood glucose level, and its concentration gradually declined over the 210 min of observation period following NNS consumption. Specifically, the estimated changes in blood glucose level relative to the baseline were 0.048 mmol/L (95% confidence interval [CI] = −0.037, 0.133) during 1–29 min, 0.006 mmol/L (95% CI = −0.177, 0.189) during 30–59 min, −0.113 mmol/L (95% CI = −0.248, 0.022) during 60–89 min, −0.041 mmol/L (95% CI = −0.194, 0.112) during 90–119 min, −0.245 mmol/L (95% CI = −0.335, −0.156) during 120–149 min, −0.108 mmol/L (95% CI = −0.170, −0.045) during 150–179 min, and −0.359 mmol/L (95% CI = −0.434, −0.283) during 180–210 min following NNS consumption. Egger’s test indicates the presence of publication bias for the estimated change in blood glucose level during the time interval of 180–210 min following NNS consumption (p-value = 0.003) but not for other time intervals.
Estimated heterogeneities in glycemic impact of NNS consumption
Meta-regressions were conducted to assess potential heterogeneities in the glycemic impact of NNS consumption by NNS type and participants’ age, BMI, and diabetic status. No difference in the glycemic impact of NNS consumption was found by NNS type for any of the seven 30-min intervals under observation (p-values > 0.05). Compared with their healthy counterparts, the changes in blood glucose level among participants with type 2 diabetes were 0.128 mmol/L (95% CI = 0.023, 0.233), 0.844 mmol/L (95% CI = 0.349, 1.338), and 0.613 mmol/L (95% CI = 0.204, 1.022) lower during 1–29 min, 150–179 min, and 180–210 min following NNS consumption, respectively. One-unit increase in BMI was found to be associated with a decrease in the glycemic impact of NNS consumption by 0.049 mmol/L (95% CI = 0.008, 0.091) during 120–149 min and 0.074 mmol/L (95% CI = 0.008, 0.140) during 180–210 min following NNS consumption. An additional year of age was found to be associated with a decrease in the glycemic impact of NNS consumption by 0.026 mmol/L (95% CI = 0.002, 0.051) during 150–179 min following NNS consumption.
Table 2 reports the results of study quality assessment. Studies included in the review on average scored 9 out of 14 with a range from 5 to 14. The distribution of qualification differed substantially across criteria. All studies clearly stated the research question, and most had a dropout rate of 20% or lower. In contrast, merely five studies [22, 26, 44, 46, 47] provided a sample size justification.
This study systematically reviewed and quantitatively synthesized existing scientific evidence from RCTs on the glycemic impact of NNSs. Twenty-nine studies with 741 total participants were identified from the keyword search of bibliographic databases. NNSs under examination included aspartame, saccharin, steviosides, and sucralose. Meta-analysis was performed to estimate and track the trajectory of blood glucose concentrations over time after NNS consumption, and test differential effects by type of NNS and participants’ age, weight, and disease status. In comparison with the baseline, NNS consumption was not found to increase blood glucose level, and its value gradually declined over the course of observation following NNS consumption. The glycemic impact of NNS consumption did not differ by type of NNS but to some extent varied by participants’ age, BMI, and diabetic status.
Findings of this review confirmed previous research on the absence of glycemic impact of NNS consumption . In addition, the review tracked the trajectory of blood glucose level over the first 210 min following NNS consumption and identified a significant decline in glycemia relative to the baseline starting at approximately 120 min. Meta-regressions further indicated that as BMI and age increased, there tended to be an inclination for decreased blood glucose levels at different time intervals. In contrast, a large cross-sectional study reported that blood glucose levels stayed constant during fasting up to 18 h . Two cross-sectional studies reported no difference in fasting blood glucose level despite increasing age  or BMI . The elderly are a vulnerable population that are exposed to an elevated risk for impaired postprandial glucose regulation. Despite normal fasting levels, as age increases the ability to regulate postprandial glycemia begins to diminish . Future studies should take these pieces of evidence into consideration and examine whether NNS consumption actually causes the decline of glycemia.
Each NNS is structurally different from the others . In addition to having a range of sweetness intensities, the structural differences lead to different amounts of NNSs absorbed and post-ingestive behaviors [55,56,57,58,59,60,61]. Despite these differences, our results suggest no differential glycemic impact by type of NNS.
The absence of glycemic impact of NNS consumption makes NNSs a potentially useful dietary aid for people with diabetes or on a weight loss regime . However, some NNS products may contain energy and carbohydrate from other sources that impact blood glucose level . Recent reviews also caution against the safety of NNS consumption on a daily basis. One prevailing hypothesis was that consuming sweet-tasting but non-caloric or reduced-calorie foods and beverages interferes with learned responses that normally contribute to glucose and energy homeostasis, resulting in the counterintuitive effect of inducing metabolic derangements . Some inconclusive evidence indicates that frequent NNS consumers could be exposed to an elevated risk for obesity, metabolic syndrome, type 2 diabetes, and cardiovascular disease . Therefore, the lack of glycemic impact may not exempt NNS from being associated with long-term adverse health consequences. Future studies are warranted to assess the health implications of frequent and chronic NNS consumption and elucidate the underlying biological mechanisms.
A recent systematic review and meta-analysis evaluated both RCTs and prospective cohort studies . NNSs were not found to be associated with decreases in BMI based on the RCTs included in the review; however, the observational data suggested that the routine intake of NNSs might be associated with an increase in BMI in the long term. Additionally, analysis of the observational studies showed an increase in cardiometabolic risks such as hypertension. However, additional experimental studies are necessary to confirm these preliminary findings.
Of the eight US Food and Drug Administration (FDA)-approved NNSs, four of them were not included in this systematic review and meta-analysis because none of the relevant studies met our inclusion criteria. Of these four NNSs (neotame, advantame, acesulfame potassium, and Luo Han Guo extract), 11 studies investigating acesulfame potassium were identified by the search algorithm [64,65,66,67,68,69,70,71,72,73,74]; however, these studies did not fit our inclusion criteria for reasons such as using an animal model  or incorporating a caloric intake in participants’ treatment . However, Bryant et al.  have shown that acesulfame potassium + glucose resulted in a 17% increase in glycemic response when compared with glucose alone, which warrants further investigation.
This review serves as the first attempt to synthesize scientific literature regarding the glycemic impact of NNSs. It included a relatively large number of RCTs with repeated clinical measures and outcomes. However, a few limitations pertaining to this review and the selected studies should be noted. Sample sizes in the included studies were small to modest, and study participants were typically non-representative to the target population. These compromised the estimation precision of the pooled effect and generalizability of the study findings. Due to the small sample size and heterogeneity in the NNSs assessed, we are not able to limit our analysis to a single NNS. The meta-analysis evaluated four NNSs, including saccharin, aspartame, sucralose, and stevia, in which sucralose and aspartame are two of the most widely used NNSs . Publication bias was identified for the estimated change in blood glucose levels during the time interval of 180–210 min following NNS consumption. This could be resulted from a small number of unrepresentative studies that followed the change in glucose level to over 3 h. Only four of the eight FDA-permitted NNSs were examined in the included studies due to data availability, so that the glycemic impact of the other NNS types remains unrevealed. No gender-specific estimate was provided in the included studies, which precluded assessing the gender difference in the glycemic impact of NNSs. Future studies are also warranted to assess the different blood glucose measurement instruments and how their sensitivity, efficiency, and convenience for the study participant may impact results.
This study systematically reviewed scientific literature regarding the glycemic impact of NNSs. Twenty-nine RCTs with 741 adult participants were identified from the keyword search. In comparison with the baseline, NNS consumption was not found to increase blood glucose level, and its value gradually declined over the course of observation following NNS consumption. The glycemic impact of NNS consumption did not differ by type of NNS but to some extent varied by participants’ age, body weight, and diabetic status. Future studies are warranted to assess the health implications of frequent and chronic NNS consumption and elucidate the underlying biological mechanisms.