Abstract
Objectives To study the associations between artificial sweeteners from all dietary sources (beverages, but also table top sweeteners, dairy products, etc), overall and by molecule (aspartame, acesulfame potassium, and sucralose), and risk of cardiovascular diseases (overall, coronary heart disease, and cerebrovascular disease).
Design Population based prospective cohort study (2009-21).
Setting France, primary prevention research.
Participants 103 388 participants of the web based NutriNet-Santé cohort (mean age 42.2±14.4, 79.8% female, 904 206 person years). Dietary intakes and consumption of artificial sweeteners were assessed by repeated 24 h dietary records, including brand names of industrial products.
Main outcomes measures Associations between sweeteners (coded as a continuous variable, log10 transformed) and cardiovascular disease risk, assessed by multivariable adjusted Cox hazard models.
Results Total artificial sweetener intake was associated with increased risk of cardiovascular diseases (1502 events, hazard ratio 1.09, 95% confidence interval 1.01 to 1.18, P=0.03); absolute incidence rate in higher consumers (above the sex specific median) and non-consumers was 346 and 314 per 100 000 person years, respectively. Artificial sweeteners were more particularly associated with cerebrovascular disease risk (777 events, 1.18, 1.06 to 1.31, P=0.002; incidence rates 195 and 150 per 100 000 person years in higher and non-consumers, respectively). Aspartame intake was associated with increased risk of cerebrovascular events (1.17, 1.03 to 1.33, P=0.02; incidence rates 186 and 151 per 100 000 person years in higher and non-consumers, respectively), and acesulfame potassium and sucralose were associated with increased coronary heart disease risk (730 events; acesulfame potassium: 1.40, 1.06 to 1.84, P=0.02; incidence rates 167 and 164; sucralose: 1.31, 1.00 to 1.71, P=0.05; incidence rates 271 and 161).
Conclusions The findings from this large scale prospective cohort study suggest a potential direct association between higher artificial sweetener consumption (especially aspartame, acesulfame potassium, and sucralose) and increased cardiovascular disease risk. Artificial sweeteners are present in thousands of food and beverage brands worldwide, however they remain a controversial topic and are currently being re-evaluated by the European Food Safety Authority, the World Health Organization, and other health agencies.
Introduction
The harmful effects of added sugars on various health outcomes including cardiometabolic disorders have been extensively studied, meta-analysed12 and are currently recognised as major risk factors by public health authorities. In particular, the World Health Organization recommends that less than 5% daily energy intake should come from free sugar.3 Artificial sweeteners emerged as an alternative to added sugar that enabled the sweet taste to be reproduced without using sugar and therefore reduced calorie content from free sugar, which was highly appreciated by consumers.4 Artificial sweeteners currently represent a $7200m (£5900m; €7000m) market globally, with a 5% annual growth projected to attain $9700m by 2028.5 An extensive number of brands worldwide contain these food additives, especially ultra-processed foods such as artificially sweetened beverages, some snacks, and low calorie ready-to-go meals or dairy products; overall more than 23 000 products worldwide contain artificial sweeteners.6 Artificial sweeteners are also directly used by consumers as table top sweeteners instead of sugar. Acceptable daily intakes for each artificial sweetener have been set by the European Food Safety Authority (EFSA), the United States Food and Drug Administration, or the Joint Expert Committee on Food Additives. Nonetheless, they remain a topic of controversy and are currently undergoing a re-evaluation by several health authorities, including the EFSA7 and WHO.8
Some experimental in vivo and in vitro studies, observational studies, and human randomised controlled trials investigated early markers of cardiovascular health, for example, weight status,9101112 hypertension,13 inflammation,14 vascular dysfunction,1516 or gut microbiota perturbation17181920 in association with consumption of artificial sweeteners or artificially sweetened beverages. Most of these studies suggested adverse effects,11121314151617181920 and few suggested neutral or beneficial properties.910 Although the results were mixed, this literature generally supports a potential involvement of artificial sweeteners in cardiovascular health, with plausible mechanisms.212223
Cardiovascular diseases (CVDs) are the leading cause of death worldwide.24 Randomised controlled trials have not directly assessed the impact of artificial sweetener intake on hard endpoints such as CVD risk for ethical reasons. Similarly, observational prospective studies have not directly investigated the association between artificial sweetener intake (mg/day) and CVD risk, but several have used artificially sweetened beverage consumption (millilitres or servings/day) as a proxy to explore these associations with conflicting results.222325262728293031323334 One of these studies was performed in the NutriNet-Santé cohort28 and found that sugary drinks and artificially sweetened beverages were associated with increased CVD risk. Systematic reviews and meta-analyses3536 have suggested direct associations between artificially sweetened beverages and CVD risk. The WHO 2022 report on the health effects of artificial sweeteners notably observed associations between consumption of beverages with artificial sweeteners (used as a proxy) and some intermediate markers of CVD,8 including a modest increase in the unfavourable total cholesterol to HDL cholesterol ratio (meta-analysis of four randomised control trials), and an increased risk of hypertension (meta-analysis of four prospective studies). The international health authority also identified an increase in CVD mortality, and in the incidence of cardiovascular events and strokes associated with greater intake of soft drinks containing artificial sweeteners (meta-analysis of four randomised control trials). However, prospective studies remain limited and the level of evidence for these associations is still considered low by WHO.8 Additionally, because artificially sweetened beverages only represent part of the total artificial sweetener intake, it is crucial to consider all dietary sources in causal studies.
In this context, our objective was to conduct a large scale prospective study using quantitative data to investigate the associations between consumption of artificial sweeteners (mg/day) from all dietary sources (beverages but also table top sweeteners, dairy products, etc), overall and by type (aspartame, acesulfame potassium, and sucralose), and risk of CVD (overall, coronary, and cerebrovascular). Our study was performed within the population based NutriNet-Santé cohort, which includes detailed information on commercial names and brands of industrial food consumed.
Methods
Study population and data collection
This study was based on the prospective NutriNet-Santé e-cohort, launched in France in May 2009, with an open ongoing enrolment of volunteers. The main objective was to investigate the relations between nutrition and health.37 Participants are French adults, aged 18 years or older, with internet access, recruited from the general population by means of multimedia campaigns. They are followed through their personal account created at inclusion on the study website (https://etude-nutrinet-sante.fr/). Immediately after enrolment, each person completes five online questionnaires about diet (24 h dietary records, detailed below), health (eg, personal and familial history, prescription drug use), anthropometric data (height and weight3839), lifestyle and sociodemographic data (eg, date of birth, sex, education level, professional occupation, smoking status, number of children40), and physical activity. Physical activity levels were defined based on the validated seven day assessment International Physical Activity Questionnaire (IPAQ).41 All activities declared by participants were converted into metabolic equivalent of task (MET) minutes per week according to the compendium of physical activities.42 Three levels of physical activity were defined: low (<600 MET-min/week), moderate (600-1500 MET-min/week), and high (>1500 MET-min/week) using standardised IPAQ processing guidelines.41 For instance, 600 MET-min/week is equivalent to 150 min/week of moderate intensity (4 METs) physical activity or 75 min/week of high intensity (8 METs) physical activity.
Each person included in the NutriNet-Santé cohort provides informed consent electronically. The study is registered at ClinicalTrials.gov (NCT03335644), conducted according to the Declaration of Helsinki guidelines, and approved by the Institutional Review Board of the French Institute for Health and Medical Research (IRB-Inserm) and the Commission Nationale de l’Informatique et des Libertés (CNIL No 908450/909216).
Dietary assessment
Three non-consecutive days of 24 h dietary records were randomly assigned over a two week period, at baseline, and every six months thereafter. During those recording days (two weekdays and one weekend day) participants indicated all foods and beverages consumed during the three main meals and any other eating occasions, and in what quantities, using validated photographs and standard serving containers43 or by directly entering the amount (in grams or millilitres). All 24 h dietary records provided during the first two years of each person’s follow-up were averaged to obtain baseline diet. This represents a reliable estimate of consumption habits, while respecting the prospective design and guaranteeing sufficient delay between consumption and CVD outcomes. Intakes of energy, alcohol, and nutrients were assessed using the NutriNet-Santé food composition table (≈3500 food/beverage items44). Nutritional contributions of mixed dishes were estimated by standard French recipes defined by nutrition professionals. Dietary assessment through these 24 h dietary records were validated against interviews by a trained dietitian45 and against blood and urinary biomarkers.4647 The basal metabolic rate and the Goldberg cut-off method enabled any under reporting to be identified484950; participants who under reported were excluded from the analyses. Supplementary method 1 gives details of methods used to identify under reporting.
Artificial sweetener intakes
Chazelas and colleagues described the quantitative evaluation of food additive consumption in participants of the NutriNet-Santé cohort.51 Briefly, food additive intakes, including artificial sweeteners, were assessed through the interactive online 24 h dietary record tool, in which commercial names and brands of industrial products consumed could be recorded. The presence of food additives was first determined for each food and beverage using ingredients lists available from three large scale food composition databases: Open Food Facts (https://world.openfoodfacts.org/)6; the French food safety agency database Oqali (https://www.oqali.fr/oqali_eng/)52; and Mintel’s Global New Products Database.53 Doses of additives were determined by around 2700 assays performed by accredited laboratories, requested by the Nutritional Epidemiology Research Team or by a consumer association (UFC Que Choisir). These quantitative data were completed by average doses per food group provided by EFSA and the Joint FAO/WHO Expert Committee on Food Additives.54 Food additive composition data were matched by date to account for possible industrial reformulations and changes in additive composition (date of consumption was considered to match the product to the closest consumption data). Supplementary method 2 gives additional information on food additive and artificial sweetener intake assessment.
For this study, we were able to estimate intakes of aspartame (European food additive identification number E951), acesulfame potassium (E950), sucralose (E955), cyclamates (E952), saccharin (E954), thaumatin (E957), neohesperidine dihydrochalcone (E959), steviol glycosides (E960), and salt of aspartame-acesulfame potassium (E962) and to create a sum variable labelled total artificial sweeteners.
Cardiovascular disease determination
Throughout follow-up, biannual health questionnaires and a permanently open personal health interface on the study account allowed participants to report any new health events, medical treatments, and examinations. For each incident CVD event declared, participants were contacted by a physician of the team and asked to provide any relevant medical records (eg, radiological reports, electrocardiogram, angioplasty). When necessary, the study physicians contacted the patient’s physician or any hospitals providing treatment to collect additional information. These medical data were reviewed by physician experts. An investigation was also conducted by the physicians of the NutriNet-Santé study by contacting the participant’s family or their physician when no connection to the study website was made for more than a year. Beyond this proactive health follow-up, data were paired with the medico-administrative databases of the national health insurance system database (SNIIRAM) and the national mortality registry (CépiDC), thereby limiting potential bias due to people with CVD not reporting their disease to the study investigators (further information available in supplementary method 3). International classification of diseases clinical modification, 10th revision, was used to classify CVD.55 For this study, first incidence of CVD, coronary heart disease (myocardial infarction, code I21; acute coronary syndrome, code I21.4; angioplasty, code Z95.8; angina pectoris, code I20.0), or cerebrovascular disease (stroke, code I64; transient ischaemic attack, codes G45.8 and G45.9) diagnosed between inclusion and 5 October 2021 were considered as events and investigated in the analyses.
Statistical analyses
Participants with at least two valid dietary records during the first two years of follow-up were included in the analysis. Those with prevalent CVD or pre-existing diabetes were excluded. To limit reverse causality bias (particularly sensitive when sugar was substituted by artificial sweeteners), participants with CVD diagnosed during the first two years of follow-up were also excluded. Supplementary figure 1 presents a flowchart showing detailed selection of the study population.
We classified participants into three categories of artificial sweetener consumption: non-consumers, lower consumers (participants with artificial sweetener intake below the sex specific median among consumers), and higher consumers. Baseline characteristics (sociodemographic, health, lifestyle, dietary intakes) were assessed for each category and compared using χ2 tests for categorical variables and analysis of variance tests for continuous variables (table 1).
Associations between artificial sweeteners, overall and the most represented (aspartame, acesulfame potassium, and sucralose, consumed by more than 5% of participants), and CVD (overall, coronary heart disease, and cerebrovascular disease) were investigated using multivariable adjusted Cox proportional hazard models (table 2). Participants contributed person time from their inclusion in the cohort until the date of CVD, date of last follow-up, date of death, or 5 October 2021, whichever occurred first. We first tested dose-response analyses using the restricted cubic spline (RCS) functions with the SAS macro developed by Desquilbet and Mariotti.56 Given the logarithmic profile of the associations suggested by the RCS curves (supplementary fig 2) and to account for the large proportion of non-consumers (especially for each specific artificial sweetener), artificial sweetener intakes were log transformed (log10 of sweetener consumption in mg/g+1) to compute continuous models (+1 was uniformly added to all consumptions because log(0) is not allowed). The continuous model was used as the primary analyses to obtain hazard ratios and 95% confidence intervals. Supplementary tables 1 and 2 presentmodels using three categories (non-consumers, lower consumers, and higher consumers, separated by the sex specific median) and four categories (non-consumers and sex specific consumers in thirds) of sweetener consumption.
The main models were adjusted for several variables suspected or known to be associated with diet and with CVD risk: sociodemographic (age, sex, educational level), lifestyle (smoking status, number of smoked cigarettes, physical activity), and health (family history of CVD) factors, and food groups and nutrients for which a role in CVD cause has been strongly suggested575859606162636465: energy intake without alcohol, alcohol, sugar, sodium, saturated fatty acids, polyunsaturated fatty acids, fibre, fruit and vegetables, and red and processed meat. We added a table showing the rationale for selection of each covariate and information on how they were collected and measured (supplementary method 4). Analyses by specific artificial sweeteners (aspartame, acesulfame potassium, and sucralose) were additionally adjusted for other artificial sweetener intakes. Multiple imputation by chained equations66 was applied to handle any missing values for covariates (15 imputed datasets; supplementary method 5). Cox proportional hazard assumption was verified using the rescaled Schoenfeld type residual method (supplementary fig 3). Competing risks were accounted for in all analyses using cause specific Cox models,67 with death considered a competing risk for CVDs, coronary heart diseases, and cerebrovascular diseases. Additionally, cerebrovascular events were considered competing risks for coronary heart diseases and vice versa. Supplementary table 3 presents results from competing events. Cumulative incidence graphs were also plotted using the Fine and Gray model (presented in supplementary fig 4).
Associations were computed separately for each type of cerebrovascular or coronary disease event: myocardial infarction, acute coronary syndrome, angioplasty, angina pectoris, stroke and transient ischaemic event (supplementary table 4), and for all CVDs except transient ischaemic events. We also investigated associations between CVD risk and artificial sweeteners from beverages and from solid food (supplementary table 5). Substitution analyses were performed by entering added sugars and artificial sweeteners into the model. Hazard ratios and 95% confidence intervals for substituting artificial sweeteners for added sugars were estimated using the difference in coefficients obtained from this model. Supplementary method 6 presents these analyses. Formal interactions between body mass index (<25 or ≥25) and artificial sweeteners were tested for each outcome by entering the product of the two variables into Cox models.
We performed a sensitivity analysis in which we doubled the requested minimal number of 24 h dietary records (excluding participants with less than four records; supplementary table 6). Additionally, we computed models with artificial sweetener intakes coded as time dependent variables across the whole follow-up period (supplementary table 6). Other sensitivity analyses were also performed, with further adjustments for prevalent dyslipidaemia, for healthy and western dietary patterns (derived by principal components analysis) instead of food groups, added sugar intakes instead of sugar, proportion of ultra-processed foods in the diet, weight loss or calorie restricted diet, weight variation during follow-up, number of 24 h dietary records, body mass index, and social desirability score68; and analyses without excluding prevalent diabetes (details presented in supplementary table 6). All tests were two sided, and P<0.05 was considered statistically significant. We used the statistical analysis software SAS, version 9.4 for analyses.
Patient and public involvement
The research question developed in this article corresponds to a concern expressed by some participants involved in the NutriNet-Santé cohort, and by the public in general. Participants in the study are thanked in the Acknowledgments section.
Results
Descriptive characteristics
Overall, 103 388 participants were selected from the NutriNet-Santé cohort. Mean age at baseline was 42.2 years (standard deviation 14.4), 79.8% were women, and the mean number of 24 h dietary records during the first two years of follow-up was 5.6 (standard deviation 3.1). Supplementary figure 5 shows the distribution of the number of 24 h dietary records per person. Among the overall cohort, 0.94% (n=1639) participants have died since their inclusion (981 in the present population study) and 9.4% (n=16 306) dropped out because they did not want to receive any more questionnaires. A total of 37.1% of participants consumed artificial sweeteners. The average intake of artificial sweeteners was 15.76 mg/day among all participants and 42.46 mg/day among consumers only, which corresponds to approximately one individual packet of table top sweetener or 100 mL of diet soda.6970 Among participants who consumed artificial sweeteners, mean intakes for lower and higher consumer categories were 7.46 and 77.62 mg/day, respectively. Compared with non-consumers, higher consumers (unadjusted comparisons) tended to be younger, have a higher body mass index, were more likely to smoke, be less physically active, and to follow a weight loss diet; they had lower total energy intake, and lower alcohol, lipid (saturated and polyunsaturated), fibre, carbohydrate, fruit and vegetable intakes, and higher intakes of sodium, red and processed meat, dairy products, and beverages with no added sugar (table 1). Aspartame, acesulfame potassium, and sucralose contributed to 58%, 29%, and 10% of total artificial sweetener intakes, respectively (fig 1). Soft drinks with no added sugar accounted for 53% of artificial sweeteners; table top sweeteners were also an important vector (30%), as well as artificially sweetened flavoured dairy products (eg, yoghurts, cottage cheese, 8%; fig 2). As shown in supplementary figure 6, food group contributions varied for each artificial sweetener; for example, table top sweeteners contributed to 48% of aspartame intake, followed by soft drinks with no added sugar (41%), whereas acesulfame potassium and sucralose were both mainly provided by the consumption of soft drinks with no added sugar (76% and 78%, respectively). Participants who consumed artificial sweeteners tended to consume more than one type of the main artificial sweeteners, and 7.23% of the total participants consumed all three of the main types (supplementary fig 7).
“>
“>