Metabolic syndrome (MetS) is characterised by a clustering of cardiometabolic risk factors within an individual, namely abdominal obesity, hypertension, and dyslipidemia [1
]. Having MetS increases the risk for cardiovascular disease (CVD), coronary heart disease, stroke, and type 2 diabetes mellitus [2
]. Modifiable lifestyle factors, such as diet, are a primary contributor to both the development and subsequent course of MetS [4
]. The average intakes of “added sugars” in the US, estimated from the National Health and Nutrition Examination Survey (NHANES), is 22 teaspoons per day [5
], of which soft drinks and fruit drinks provide more than 40% [6
]. Consumption of soft drinks is increasing amongst men, women and children in the US and also in Europe [7
]. This may include both sugar-sweetened ‘regular’, and artificially sweetened ‘diet’ soft drinks. Substantial epidemiological evidence for the association between high intakes of sugar-sweetened beverages and risk for obesity, type 2 diabetes, and CVD exists [11
]. While replacing sugar-sweetened soft drinks with diet, sugar-free, or artificially-sweetened beverages may be used to reduce sugar intake, recent research has demonstrated associations between diet soft drink consumption and adverse cardiometabolic outcomes [16
Diet soft drink intakes have been positively associated with MetS prevalence in the Framingham Offspring Study [16
], and with greater relative risk of incident MetS in the Multi-Ethnic Study of Atherosclerosis (MESA) [19
], and in the Atherosclerosis Risk in Communities Study (ARIC) [18
These studies have all been conducted in the US [16
]. Less evidence for detrimental health effects associated with high soda consumption has emerged from Europe. One study has shown a positive association between sugar-sweetened beverages and six-year risk of MetS in a Mediterranean cohort of university graduates [20
]. The Oslo Health Study showed a positive association between a diet index reflecting a high intake of soft drinks (including sugar-free) and a low intake of fruit and vegetables with components of MetS [21
]. Further research in European samples is warranted.
The objective of the present study was to evaluate soft drink intakes from two cross-sectional study samples in central New York (Maine-Syracuse Longitudinal Study, MSLS, USA), and Luxembourg (Observation of Cardiovascular Risk Factors in Luxembourg, ORISCAV-LUX), and to assess intakes in relation to the prevalence of MetS and its individual components. Comparisons between two male cohorts from America and Italy have been made in relation to cardiovascular health risk factors and alcohol intake [22
]. To our knowledge, no such comparisons have been made for soft drink consumption, in relation to cardiovascular health, and MetS in particular.
In addition, we aimed to determine whether the prevalence of MetS differed according to the type of soft drink consumed (regular versus
diet). In our past examination of the health and lifestyle habits in these two studies, we found that the ORISCAV-LUX participants had healthier diets and lower levels of obesity, hypertension, diabetes and CVD than those in MSLS [23
]. The following hypotheses were therefore advanced: (1) overall soft drink consumption would be higher in the MSLS than in ORISCAV-LUX; (2) positive relations would be found between both regular and diet soft drink intakes and MetS prevalence in both studies; and (3) modestly stronger relations would be observed in the MSLS.
2.1. Study Design and Sample
The total sample comprised from two studies, the MSLS, and the ORISCAV-LUX, included 2126 individuals, 803 from MSLS and 1323 from ORISCAV-LUX. Further details related to the methods of sampling for both studies appear below and in numerous publications [24
2.1.1. Participants in MSLS (USA)
The MSLS is a longitudinal, community-based study of aging, cardiovascular risk factors and cognitive functioning in adults, aged 23–98 years [26
]. The MSLS was conducted in Syracuse, New York (NY), USA and its catchment area (Central NY). At initial recruitment (1975), the sole exclusions were institutionalized people, diagnosed alcoholism and psychiatric disorder. The data for the present cross-sectional study were taken from subjects returning for the sixth (2001–2006) study wave when dietary intake measures were first obtained and data on objectively measured cardiovascular risk factors were available. Beginning with a sample of 1049 individuals, participants were excluded from the present analysis for the following reasons: missing data on diet or components of MetS (n
= 169), acute stroke (n
= 28), probable dementia (n
= 8), undergoing hemo-dialysis (n
= 5), inability to read English (n
= 1), and alcohol abuse after baseline (n
= 1), leaving 803 participants.
The University of Maine Institutional Review Board approved this study and informed consent was obtained from all participants.
2.1.2. Participants in ORISCAV-LUX (Luxembourg)
ORISCAV-LUX was a nationwide, cross-sectional study on the prevalence of cardiovascular risk factors among the adult population of Luxembourg, aged 18–69 years, conducted in 2007–2009. Exclusions were pregnancy (n
= 21), serious mental and/or physical handicap (n
= 5), prisoners (n
= 1), people outside the determined age range (n
= 2) and those deceased before recruitment (n
= 5) [25
]. A representative random sample of 1432 individuals, stratified by sex, age, and district of residence completed the recruitment procedure [24
]. After data cleaning, the total ORISCAV-LUX sample comprised 1323 individuals.
The study was approved by the National Research Ethics Committee and the National Commission for Private Data Protection, and all participants gave informed written consent.
2.2.1. Dietary Assessment
In the MSLS, dietary intake was assessed using the food frequency questionnaire (FFQ) component of the Nutrition and Health Questionnaire [30
]. Its acceptable validity has been demonstrated by comparison with dietary recall, protein excretion and total energy expenditure data [30
]. The dietary component questions participants about their frequency of consumption of 37 foods and beverages. Participants are required to stipulate their frequency of consumption, with six response options: “never”, “seldom”, “once a week”, “2–4 times/week”, “5–6 times/week” and “once or more/day”. For soft drinks, participants were asked to report how many glasses/cans of “diet” carbonated soft drinks and “regular” carbonated soft drinks they consumed daily. These two intakes were summed to give total soft drink intake per day. Portion or serving sizes were not stipulated; therefore total energy was estimated in the following manner: the median score within each response option was used to estimate total intakes per week; for example, two to four times per week was estimated at three. The mean number of times each food was consumed on a weekly and then daily basis was calculated for all foods. Individual foods were categorized into six major food groups (grains, fruits, vegetables, protein foods, dairy foods, and fats/sweets/other). Total energy was therefore estimated by summing the number of servings per day of all foods and beverages [33
In ORISCAV-LUX, dietary intake was assessed using a validated, semi-quantified FFQ, which assessed the frequency of consumption of 134 items over the previous three months [34
]. Participants were asked how frequently they consumed one standardized portion of each food. Both diet and regular soft drinks were included in the FFQ, and for beverages, there were five frequency response categories: “never or rarely”, “1–3 times/month”, “1–2 times/week”, “3–5 times/week”, and “every day”. Participants were also required to indicate the total quantity (in mL) of drink they consumed each time. This enabled the calculation of daily intakes of diet, regular, and total soft drinks (in servings per day, 330 mL equating to one serving). Energy and nutrient intake data, including alcohol (g/day) and total energy intake (Kcal/day), were compiled.
2.2.2. Lifestyle and Heath Data
Participants in both studies underwent physical and anthropometric measurements, and blood tests. Standardized protocols for data collection were used. Body weight, height, body mass index (BMI), waist circumference, and blood pressure (BP) measures were assessed as described previously for both studies [24
]. Standard assay methods were employed [26
] to obtain fasting plasma glucose, serum triglycerides, and HDL-cholesterol, as well as low-density lipoprotein (LDL)-cholesterol and total cholesterol (all mg/dL).
All participants completed self-administered questionnaires to gain information on demographic and socioeconomic characteristics, including age, sex, education (years), smoking (cigarettes/day), and physical activity (minutes/day). In the MSLS, physical activity was measured with the Nurses’ Health Study (NHS) Activity Questionnaire [37
]. Smoking status was based on self-report from the Nutrition and Health Questionnaire [30
]. In ORISCAV-LUX, physical activity was measured using the short format International Physical Activity Questionnaire (IPAQ) [38
]. Detailed data regarding smoking were obtained from the health questionnaire.
2.2.3. Definition of MetS
MetS (and components) was defined by National Cholesterol Education Program Adult Treatment Panel III criteria. MetS was defined as present if three out of five risk factors were present: waist circumference ≥ 88 cm for women or 102 cm for men; fasting glucose ≥ 100 mg/dL (or drug treatment); blood pressure ≥135/85 mmHg or treatment for hypertension; serum triglycerides ≥150 mg/dL (or drug treatment); and high density lipoprotein (HDL)-cholesterol ≤ 40 mg/dL in men or 50 mg/dL in women [39
2.3. Statistical Analysis
Participant characteristics in each study were compared according to soft drink consumption: non-consumer, and three consumer groups (diet, regular, or mixed diet/regular). A consumer was defined as someone who reported consuming any type of soft drink. For continuous variables, analysis of variance (ANOVA) was used to compare the four groups in terms of demographics, health factors, and dietary variables, with Bonferroni adjustment for multiple comparisons. For categorical health-related variables, Chi square tests were performed. For the primary analyses, logistic regression analyses were used to compare the prevalence of MetS in participants who consumed soft drink (one per day, two or more per day), compared to non-consumers (referent group). This was performed for diet, regular, and total soft drinks. The same analyses were performed for the MSLS (n = 803) and for ORICAV-LUX (n = 1323). The following three multivariable regression models were used:
Model 1: adjusted for demographic and lifestyle factors, including age, sex, education, smoking, and physical activity.
Model 2: Model 1 plus adjusted for dietary factors including intakes of alcohol, vegetables, fruit, grains and meat.
Model 3: Models 1 and 2 plus adjusted for total energy intake.
When assessing relations between diet soft drink and MetS prevalence, regular soft drink intake (servings/day) was included in Models 2 and 3; similarly diet soft drink intake was added to Models 2 and 3 when assessing associations between regular soft drink intake and MetS.
Finally, multiple linear regression analyses were used to evaluate relations between soft drink consumption and each of the individual components of MetS, as continuous variables (waist circumference, systolic and diastolic BP, HDL-cholesterol, triglycerides, and fasting plasma glucose), in each study. The same covariable sets were used as for the logistic regression analyses (see above).
All statistical analyses were performed with PASW for Windows® version 21.0 software (formerly SPSS Statistics Inc. Chicago, IL, USA); p < 0.05 was considered statistically significant.
Consistent with previous research [16
], we observed similar, significant associations between increasing soft drink consumption and prevalence of MetS in Central NY, USA and in Luxembourg. In particular, associations between diet soft drink intakes and MetS prevalence were observed. Consistent with these findings, diet soft drinks were positively associated with waist circumference.
These findings were consistent in the two studies, despite quite different soft drink intake patterns. Further, MetS prevalence was considerably higher in the US (Central NY) sample (44%) than in the Luxembourg sample (26%). These data are consistent with a previous comparative analyses of the cardiovascular health of two samples from these studies [23
]. Based on these previous findings, and contrary to what we expected, more people consumed soft drinks in ORISCAV-LUX than in MSLS, but the quantities consumed were higher in MSLS. Although we hypothesised that associations may be stronger in MSLS than in ORISCAV-LUX, the odds of having MetS associated with total soft drink intakes were similar in both (approximately 2-fold higher odds). Of soft drinks consumers in the US sample, a greater proportion selected diet drinks over regular/sugar-sweetened soft drinks; while the opposite was true in ORISCAV-LUX, i.e.
, the proportion of those consuming regular soft drinks was three times higher than those consuming diet drinks. Approximately 2% of the ORISCAV-LUX sample consumed two or more diet soft drinks per day, compared with 10% in the MSLS sample. Interestingly, the MSLS sample was more overweight and centrally obese than the Luxembourg one (as measured by both BMI and waist circumference, respectively), despite fewer persons consuming soft drinks. The age difference in the two samples may also help to explain these observations, as the MSLS participants consisted of a greater number of older adults, who typically consume fewer soft drinks than younger people [9
]. In individuals who did consume soft drink, the average daily intake of soft drink consumers (any type) in the MSLS sample was over two times greater than the intake of consumers in ORISCAV-LUX (1.8 ± 1.6 servings/day in MSLS, compared to mean 0.8 ± 1.3 servings/day in ORISCAV-LUX). When we equated the distributions of the two samples (to the common age range of 23 to 69 years) in a secondary analysis, the mean number of diet, regular, and total soft drinks consumed per day (amongst soft drink consumers) remained higher in Central NY than in Luxembourg, when adults aged over 69 years were excluded (see Supplementary Table S1
Other recent studies have demonstrated that higher levels of soft drink consumption (at least daily intakes) are associated with MetS [16
]. In MESA [19
], at least daily consumption of diet soda was associated with a 36% greater relative risk of incident MetS, compared with non-consumption. Of the components, waist circumference and high fasting glucose were prospectively associated with diet soda consumption. The present study confirms these findings, at the level of one daily serving. Increasing to two daily servings did not significantly increase the likelihood of having MetS; in both studies proportions of those with MetS were similar regardless of whether one drink per day, or more than one, was consumed. Furthermore, and consistent with MESA, we showed an increasing dose-response pattern in both samples for waist circumference and in ORISCV-LUX for fasting plasma glucose and systolic BP, with both increasing as more diet soft drink was consumed.
A number of mechanisms have been postulated that may explain the findings observed. Firstly, those who consume higher quantities of soft drinks may also have a dietary and/or lifestyle pattern that is not as healthy as those who do limit these drinks. Total energy intakes were higher in those individuals who consumed regular soft drinks compared to those who consumed diet drinks in both samples. Regular soft drink consumers in the studies analysed here (both MSLS and ORISCAV-LUX) consumed fewer fruit and vegetables, more grains and meats, and smoked more cigarettes than diet soft drink consumers. Findings in a previous study suggest people who consume higher amounts of sugar-containing soft drinks may fail to compensate for these ‘liquid calories’ at their next meal, promoting a positive energy balance and weight gain [40
]. The energy compensation made for beverages is not equivalent to that made for solid foods, and therefore the energy content of soft drinks can contribute to a cumulative excess of energy over time to produce obesity [41
]. The high fructose corn syrup added to regular soft drinks (the primary sweetener in soft drinks) may also contribute to adverse metabolic effects. Less is known about the physiological mechanisms linking high intakes of sugar-free soft drinks with adverse cardiometabolic outcomes. Animal studies have shown that artificial sweeteners, such as aspartame, may reduce the ability of the body to estimate the energy content of foods, leading to increased intake and body weight gain [42
]. However, the safety of aspartame, for use as a sweetener and flavour enhancer, has been established [43
]. Some researchers have suggested that the high sweetness in artificially sweetened drinks may result in hunger [44
] or greater preference for other sweet or energy dense foods [42
]. However other studies have failed to show that artificial sweeteners (including both aspartame and saccharin) increase hunger or subsequent food intake [45
]. Other research suggests that positive relationships may be due to confounding or reverse causality [48
]. For example, diet soda consumption has been reported as up to three times higher among adult diabetics in the US than non-diabetics [49
]. People diagnosed with heart disease or diabetes may therefore actively opt for artificially sweetened drinks.
Indeed reverse causation may explain the present findings between diet soft drink consumption and higher MetS prevalence. It is possible that some current drinkers of diet soft drinks had replaced regular with diet drinks for health reasons, and therefore continued to exhibit adverse disease patterns. This may be particularly true within the Luxembourg sample, where levels of type 2 diabetes, hypertension and obesity (BMI ≥ 30 kg/m2) were higher in diet drinkers than in regular soda drinkers. Excluding those being treated for diabetes (both studies) did not however alter the results. Of note in both samples, was the observation that obesity levels in diet soft drink consumers was significantly higher than in non-drinkers (of any type). It is quite plausible that in response to their body weight status, some individuals may have switched from regular to diet drinks, but not given up soft drinks altogether, explaining the higher obesity levels in this group. However, this cannot fully explain the findings as excluding those who were dieting in the Luxembourg sample did not alter the significant positive associations between diet soft drink consumption and MetS.
Strengths and Weaknesses
Strengths of the analysis include detailed information in both MSLS and ORISCAV-LUX on diet, cardiometabolic health, and additional covariates in adults. This is the first study that we are aware of to compare relationships between soft drink consumption and MetS in two studies from two different countries. Cross-country comparisons are important to provide insights into the social determinants of dietary habits and health [50
There are several study limitations. The ORISCAV-LUX was a nationwide, population-based study, whereas MSLS was a community-based sample restricted to Central NY, USA. A broader American sample would enable us to see if there are similar or differing trends in other parts of the US. The cross-sectional nature of both studies prohibits any conclusions with regard to causality. Soft drink intakes and other dietary data were based on participant self-report and the same food questionnaires were not used in both studies. The FFQ used in ORISCAV-LUX was semi-quantitative, with participants reporting frequency of servings and stipulating their serving size in mL. The MSLS participants reported their intakes in terms of glasses or cans per day. The inherent measurement error associated with the use of FFQ’s should also be acknowledged. Further, validation studies for the Nutrition and Health Questionnaire (used in the U.S. sample) have been performed in European samples. Confounding by other dietary or lifestyle factors can also not be ruled out. We have statistically adjusted for a number of variables that related to both predictor (soft drink intake) and outcome (MetS) in both studies, however there may be other unknown factors which impact the relations observed. It also must be acknowledged that the two studies were not conducted at the same period of time, with the MSLS data being collected approximately five years prior to ORISCAV-LUX. One may hypothesise that the availability of diet soft drinks may have increased over time (e.g., more varieties, availability, and accessibility), particularly in Europe. However, our examination of this data actually shows higher consumption of diet drinks in MSLS than in ORISCAV-LUX; whereas the opposite may have been observed if diet consumption was notably higher in 2007-09 than in 2001-06. The time period difference is unlikely to have impacted upon the study findings.
While we are not attempting to generalise beyond the two geographic study sites, the present study does provide insight into how cardiometabolic health differs between the two locations and demonstrates the robust nature of our findings of an association between soft drink consumption and MetS.