July 31, 2007
assessed the dietary information on consumption of total calories, saturated fat, trans fat, fiber, magnesium, and glycemic index from the FFQ.19 Because a FFQ was not administered at the fourth examination cycle, dietary covariate data from the fifth examination cycle were used for analyses using information from the examination cola questionnaire at all 3 examinations.
Data from the FFQ were considered valid only if total energy intakes reported were 2.51 MJ/d (600 kcal/d) for men and women but 17.54 MJ/d (4200 kcal/d) for men or 16.74 MJ/d (4000 kcal/d) for women and if fewer than 13 food items were left blank. Each food item was categorized in 9 categories that ranged from never or 1 serving per month to 6 servings per day. For assessment of saturated fat, trans fat, or dietary fiber, the nutrient intakes from all specific food items were multiplied by the frequency of consumption. The validity of the FFQ has been demonstrated previously.21
Definition and Components of the Metabolic Syndrome
The metabolic syndrome was considered present if 3 or more of the following individual components were present14,22: waist circumfer- ence 35 inches (88 cm) for women or 40 inches (102 cm) for men; fasting blood sugar 100 mg/dL (5.5 mmol/L) or treatment with oral hypoglycemic agents or insulin; blood pressure 135/ 85 mm Hg or treatment for hypertension; serum triglycerides 150 mg/dL (1.7 mmol/L) or treatment for hypertriglyceridemia (with niacin or fibrates); and HDL-C 40 mg/dL (1.03 mmol/L) in men or
50 mg/dL (1.3 mmol/L) in women.
Age- and sex-adjusted baseline characteristics of the participant groups defined according to the number of soft drinks consumed in 1 day (1, 1, or 2 per day) were compared by multiple linear and multiple logistic regression analysis for continuous and categorical characteristics, respectively. Data on consumption of soft drinks at each of the 3 eligible baseline examinations (examination cola questionnaire) were used for this purpose. Tests for trend in baseline characteristics across soft drink consumption categories were per- formed with multiple regression. We also assessed the baseline characteristics after excluding participants with prevalent metabolic syndrome at baseline examinations (sample used for incidence analyses; see below).
Soft Drink Consumption and Prevalence of the Metabolic Syndrome
We used data from examinations 4, 5, and 6 (examination cola questionnaire) and generalized estimating equations to compare the prevalence of metabolic syndrome in participants who consumed 1 soft drink per day with those who consumed 1 soft drink per day (referent). Each participant could contribute up to 3 person- examinations of data for analysis. We also evaluated a dose response by comparing individuals who consumed 1 soft drink per day and those who consumed 2 soft drinks per day with the referent group. We constructed multivariable models in hierarchical fashion with adjustment for age and sex (model I) and for age, sex, physical activity index, smoking, dietary consumption of saturated fat, trans fat, fiber, magnesium, total calories, and glycemic index (model II).
We used soft drink consumption data from FFQs at examinations 5 and 6, which yielded a smaller sample (Figure), to relate the prevalence of metabolic syndrome across the following categories of intake of regular versus diet soft drinks using generalized estimating equations: (1) 1 diet or regular soft drink per week (referent), (2) 1 to 6 diet soft drinks per week, (3) 1 diet soft drink per day, (4) 1 to 6 regular soft drinks per week, (5) 1 to 6 regular or diet soft drinks per week, and (6) 1 regular soft drink per day. Individuals reporting consumption of both diet and regular soft drinks 1/d (n16) were grouped into the last category empirically. We evalu- ated the 2 sets of models (I and II) noted above.
Downloaded from http://circ.ahajournals
Soft Drink Consumption and Incidence of the Metabolic Syndrome
To assess the relations of soft drink consumption to the incidence of metabolic syndrome, we excluded participants with prevalent metabolic syndrome at each of examination cycles 4, 5, and 6 (n2897 person-observations). Then, we used pooled logistic regression analyses by combining each 4-year follow-up period of observations to relate the number of soft drinks consumed per day (examination cola questionnaire) to the incidence of metabolic syndrome (from examination cycles 4 to 5, 5 to 6, and 6 to 7).23 The eligible participants were free of metabolic syndrome at each baseline examination, and in this setting, pooled logistic regres- sion has been shown to provide risk estimates similar to time- dependent Cox models.24 We compared the consumption of soft drinks 1 per day with infrequent drinkers (1 per day; referent) and also tested for a dose response by comparing groups consum- ing 1 and 2 soft drinks per day with the referent group. We evaluated 2 sets of models (covariates as in models I and II above), which paralleled the analyses of prevalence of metabolic syndrome.
Consumption of soft drinks varies with age and by sex.25 It has also been suggested that the effects of soft drinks and carbohydrates on metabolic traits may vary according to age, sex,26 and baseline body weight.27 Therefore, we assessed for effect modification by age (modeled as a continuous variable), sex, and body mass index (30 versus 30 kg/m2) by incorporating appropriate interaction terms in the multivariable models. We repeated analyses with additionally adjustment for alcohol consumption and baseline levels of systolic and diastolic blood pressure, blood glucose, serum triglycerides, and HDL-C. These models were constructed to account for baseline levels of metabolic traits. Additionally, we repeated analyses to examine the association between consumption of caffeinated and decaffeinated soft drinks, considered separately, and incidence of the metabolic syndrome. Because individuals with diabetes mellitus are a particularly high-risk group for developing metabolic abnormali- ties, we also repeated our analyses after excluding those with prevalent diabetes mellitus at baseline.
To compare the risk of new-onset metabolic syndrome according to the type of soft drink consumed (regular versus diet), we used data from the FFQs at examinations 5 and 6 and evaluated the incidence of the metabolic syndrome across categories of soft drinks con- sumed. The 6 categories of regular and diet soft drinks were those noted above (for the analyses of the prevalence of metabolic syndrome), and 2 sets of models were evaluated (models I and II, as described above).
Incidence of Individual Components of Metabolic Syndrome
We used multivariable logistic regression to evaluate the relations of soft drink consumption to the incidence of each individual compo- nent of metabolic syndrome using data from the examination cola questionnaire. We excluded participants who had the specific meta- bolic trait prevalent at baseline; for example, we excluded individ- uals with blood glucose 100 mg/dL (5.5 mmol/L) from the “at-risk” group for analysis that examined the incidence of impaired fasting glucose. Thus, we examined the incidence of increased waist circumference, impaired fasting glucose, high blood pressure, hyper- triglyceridemia, and low HDL-C (all defined as above) according to the number of soft drinks consumed per day.
We evaluated 2 sets of models (I and II, as noted above) and compared the risk of developing metabolic traits associated with consumption of 1 soft drinks per day with that in infrequent drinkers (1 soft drinks per day). We also evaluated for a dose response as detailed above. We did not perform analyses of devel- opment of individual metabolic syndrome components in relation to regular versus diet soft drink intake using the FFQ data at examina- tions 5 and 6 because the grouping of incident events into 6 categories resulted in modest numbers of events in each category.
.org/ by guest on February 4, 2015