Cumulative sugar-sweetened beverage consumption is associated with higher concentrations of circulating ceramides in the Framingham Offspring Cohort

ABSTRACT

Background

Ceramides have been implicated in the pathogenesis of type 2 diabetes and cardiovascular disease. Limited data exist on how habitual dietary intake of foods that can alter hepatic lipid metabolism may influence circulating ceramide concentrations.

Objectives

We investigated the cross-sectional association of cumulative sugar-sweetened beverage (SSB) consumption with concentrations of 3 circulating ceramides and ceramide ratios.

Methods

We examined participants from the Framingham Heart Study's Offspring Cohort who had 3 ceramides measured (n  = 1561, mean age 66 y, 59% women). SSB consumption was measured 4 times over ∼14 y. Participants were categorized by cumulative SSB intake as nonconsumers (0 to <1 SSB serving/mo) and occasional (1 SSB serving/mo to <1 serving/wk), frequent (1 SSB serving/wk to <1 serving/d), and daily (≥1 SSB serving/d) consumers. Multivariable linear regression models were used to relate cumulative SSB consumption (independent variable) to blood concentrations of ceramides (C16:0, C22:0, and C24:0) and ceramide ratios (C22:0/C16:0 and C24:0/C16:0).

Results

In adjusted models, more frequent cumulative SSB consumption was positively associated with concentrations of the C16:0 and C22:0 ceramides (P_trend < 0.05). Compared with nonconsumers, daily consumers had 0.01 μg/mL (95% CI: 0.002, 0.017 µg/mL) and 0.06 µg/mL (95% CI: 0.018, 0.092 µg/mL) higher mean concentrations of the C16:0 and C22:0 ceramides, respectively. Results were consistent when modeling continuous cumulative SSB consumption per 1 serving/d. We observed effect modification by diabetes status in the relation between cumulative SSB consumption and concentrations of the C24:0 ceramide (P_interaction = 0.014). In a stratified analysis, more frequent cumulative SSB consumption was positively associated with concentrations of the C24:0 ceramide only in individuals with prediabetes or diabetes (P_trend = 0.001).

Conclusions

Our study raises the possibility that higher concentrations of distinct ceramide species, previously associated with adverse metabolic health, may be one mechanism by which SSB consumption contributes to higher risk of cardiometabolic diseases.

Introduction

Ceramides are bioactive sphingolipids characterized by a sphingoid base and an acyl chain of varying length (2). Ceramides are synthesized de novo from the condensation of serine and palmitate (2). This unique class of lipids has been implicated in lipotoxicity, which is thought to promote metabolic abnormalities often associated with obesity (3). Research in rodent models implies a direct role for total ceramides in the underlying pathologies of type 2 diabetes mellitus (T2D) and atherosclerosis (3). Emerging evidence suggests that concentrations of distinct circulating ceramides in humans may serve as biomarkers of insulin resistance, T2D, and cardiovascular disease (CVD) (4). In particular, ratios of long-chain (C14–C18) and very-long-chain (≥C20) ceramides to each other may provide incremental information for CVD risk prediction beyond standard risk factors (5–7). It has previously been demonstrated that over-nutrition and excess consumption of dietary SFAs are associated with higher concentrations of circulating ceramides (8, 9). However, the relations between individual dietary components, at habitual intake amounts, and circulating ceramide concentrations are less clear.

Sugar-sweetened beverages (SSBs) are widely recognized as key dietary risk factors for obesity and poor cardiometabolic health (10, 11). Longitudinal studies indicate that frequent SSB consumption is associated with a higher risk of T2D and CVD (12–14). In the United States, SSBs account for nearly half of all added sugar consumed and are the major source of dietary fructose (13). It remains unclear if the detrimental effects of SSB consumption are primarily due to excess calories, are unique to excess added sugar and/or fructose, or are in fact due to other differences in overall diet quality. Fructose metabolism interconnects with hepatic lipid metabolism by the ability of fructose to serve as a substrate for de novo lipogenesis (DNL) and regulate lipogenic gene expression in the liver (15, 16). Thus, excess SSB consumption could influence de novo ceramide synthesis by way of persistent alterations in hepatic DNL.

We aimed to investigate the cross-sectional association between habitual cumulative consumption of SSBs and concentrations of 3 circulating ceramide species (C16:0, C22:0, and C24:0). Because ceramide ratios have emerged as sensitive predictors of CVD mortality, we also investigated the association between SSB consumption and a priori defined ceramide ratios (C22:0/C16:0 and C24:0/C16:0). We hypothesized that more frequent consumption of SSBs would be directly associated with higher concentrations of circulating ceramides but not with ceramide ratios, which would imply higher concentrations of the C22:0 and C24:0 ceramides than of the C16:0 ceramide.

Methods

Study sample

The Framingham Offspring Cohort was initiated in 1971 with the enrollment of 5124 adults and has been described elsewhere (17). Beginning in 1971, enrolled participants have undergone a physical assessment at examinations that occur approximately every 4 y (examination cycles). We evaluated data from participants of the Framingham Offspring Cohort who attended the eighth examination cycle (2005–2008). Blood concentrations of 3 ceramides were previously measured in 2843 of the 3012 participants at the eighth examination cycle (6) and these participants were eligible for the present investigation. Of these 2843 eligible individuals, we excluded 701 for the following reasons: 312 participants with missing dietary data, 337 with prevalent CVD (stroke or myocardial infarction), and 52 with missing other covariate data. We then excluded participants who did not have complete SSB data across the fifth, sixth, seventh, and eighth examination cycles (n = 581). The final sample size for our cross-sectional analysis was 1561 participants (Supplemental Figure 1). The Boston University Medical Center and Tufts University Health Sciences Institutional Review Boards approved the study protocol and all participants provided written informed consent.

Quantification of ceramides

For our primary outcome of circulating ceramides, we used measures of 3 ceramides (C16:0, C22:0, and C24:0) that had previously been quantified (in micrograms per milliliter) at the eighth examination cycle of the Framingham Offspring Cohort (6). For our secondary outcome, we examined ratios of very-long-chain and long-chain ceramides to each other (C22:0/C16:0 and C24:0/C16:0). The quantification of plasma C16:0, C22:0, and C24:0 ceramides, by a validated LC/tandem MS assay, in examination cycle 8 of the Framingham Offspring Cohort has been previously described at length (6). In brief, plasma samples (50 µL) were spiked with internal standards containing known amounts of the respective ceramides. Lipid extraction was completed using a 9:1 isopropanol:chloroform solution. Ceramides were analyzed on a Shimadzu HPLC system coupled with an Applied Biosystems/MDS Sciex 4000QTRAP mass spectrometer using multiple reaction monitoring. Analyst software (version 1.5.2, SCIEX) was used for data acquisition and analysis. The CV was 7.8%, 7.6%, and 6.9% for the C16:0, C22:0, and C24:0 ceramides, respectively (6).

Dietary assessment

The Harvard semiquantitative FFQ was used to assess SSB consumption (18). The Harvard FFQ contains 126 food items with standard serving sizes and 9 frequency categories to indicate usual consumption over the past 12 mo. Frequency categories ranged from none or <1 serving/mo to ≥6 servings/d. The Harvard FFQ has previously been validated against 7-d dietary records in the Health Professionals Follow-up Study (18, 19). The Pearson correlation coefficients from a prior assessment of the Harvard FFQ SSBs and 7-d dietary records were 0.84 for colas (caffeinated and caffeine free), 0.36 for other carbonated beverages, and 0.56 for noncarbonated fruit drinks (20). We considered individual FFQs to be valid if there were <13 blank items and estimated daily caloric intake was ≥600 kcal/d and <4000 kcal/d for women or <4200 kcal/d for men. On the FFQ, SSBs included 1) caffeinated colas with sugar; 2) caffeine-free colas with sugar; 3) other carbonated beverages with sugar; and 4) fruit punches, lemonade, or other noncarbonated fruit drinks. Beverages containing nonnutritive sweeteners (diet beverages) included 1) low-calorie caffeinated colas; 2) low calorie caffeine-free colas; and 3) other low-calorie carbonated beverages.

Anthropometry and covariate assessment

Participants underwent a physical examination and medical history using standard protocols at each Framingham Heart Study visit. We assessed all lifestyle, dietary, and adiposity variables at examination cycle 8. We calculated nutrient and energy intake from the aforementioned FFQ. Physical activity status was calculated based on time and intensity of activities in a day (21). Diet quality was determined using the 2010 Dietary Guidelines Adherence Index (DGAI) (22). We defined current smokers as participants who smoked regularly in the year preceding examination cycle 8. Use of lipid-lowering medications was based on self-reported use for treatment of hypercholesterolemia in the past year. BMI was calculated as kg/m². Waist circumference (in inches) was measured at the level of the umbilicus. Individuals with fasting plasma glucose concentrations ≥5.6 mmol/L (100 mg/dL) or who self-reported use of antidiabetic medications were considered prediabetic/diabetic.

Statistical analysis

Averaging across repeated dietary assessments better captures the long-term dietary intake and reduces potential error due to misclassification (23). We calculated cumulative SSB consumption as the mean consumption, among participants who had complete SSB data from the fifth (1991–1995), sixth (1995–1998), seventh (1998–2001), and eighth (2005–2008) examination cycles (n = 1561), representing usual intake over 14 y. We then categorized participants based on frequency of cumulative SSB consumption as follows: nonconsumers (0 to <1 SSB serving/mo), occasional consumers (1 SSB serving/mo to <1 serving/wk), frequent consumers (1 SSB serving/wk to <1 serving/d), and daily consumers (≥1 SSB serving/d).

Pearson correlation coefficients between individual ceramides were calculated adjusting for age, sex, and use of lipid-lowering medication. Multivariable linear regression models were used to relate frequency of cumulative SSB consumption (independent variable) to concentrations of circulating ceramides and ceramide ratios (C16:0, C22:0, C24:0, C22:0/C16:0, and C24:0/C16:0; dependent variables, separate model for each) at examination cycle 8. We completed analyses examining cumulative SSB consumption as a categorical variable and as a continuous variable in multivariable linear regression models. When modeling by SSB consumption category, β coefficients and 95% CIs are reported with respect to nonconsumers as the referent group. To test for a linear trend across SSB consumption categories, we assigned the median SSB intake of a consumption category as the intake to all individuals in the respective consumption category. A significant linear trend was then tested using the median intakes across categories as continuous values in linear regression models. P values from the test for linear trend (P_trend) are presented with least-square means across SSB consumption categories. We also repeated regression analyses with cumulative SSB consumption modeled as a continuous exposure.

Model 1 adjusted for age (years), sex, smoking status, use of lipid-lowering medication, total energy intake, DGAI score, alcohol intake, and physical activity. This model was repeated with the inclusion of BMI (continuous). We tested for potential effect modification of the association between cumulative SSB consumption and circulating ceramides (C16:0, C22:0, and C24:0) by the use of lipid-lowering medication, BMI (<25 compared with ≥25), and diabetes status (prediabetic/diabetic compared with nondiabetic). Cross-product terms for the respective dichotomous variables were included in multivariable regression models testing for linear trend between frequency of SSB consumption and concentrations of ceramides (C16:0, C22:0, and C24:0). The significance level for the interactions was set at P ≤ 0.017 (0.05/3) to account for multiple tests.

SSB intake contributes to the added sugar component of the DGAI. Thus, we completed a sensitivity analysis using a modified DGAI, with no added sugar component, as a covariate. In an additional sensitivity analysis, we included participants with prevalent CVD and included CVD status as a covariate in our multivariable models. We also repeated our multivariable analyses adjusting for the same covariates as aforementioned but including further adjustment for the consumption of diet beverages. Recent evidence suggests that distinct ceramide species may be associated with visceral adiposity (24). Hence, analyses were also repeated using waist circumference in place of BMI. Lastly, we conducted analyses in a sample restricted to SSB consumers only (occasional, frequent, and daily).

For all analyses, we considered a 2-tailed P < 0.05 statistically significant, except where otherwise stated. The statistical analysis was completed using SAS statistical software version 9.4 (SAS Institute).

Results

Study sample characteristics

Participant characteristics were calculated across consumption categories and are displayed as mean ± SD, or frequency and proportion. Unadjusted concentrations of the C16:0, C22:0, and C24:0 circulating ceramides; ceramide ratios; and characteristics of the study population across categories of cumulative SSB consumption are displayed in Table 1. Most participants (44%) were frequent SSB consumers, whereas ∼28% of participants were nonconsumers, ∼20% were occasional consumers, and ∼8% were daily consumers. Daily SSB consumers were younger, more likely to be men, more likely to be a current smoker, were slightly more physically active, consumed more energy, and had a lower diet quality as assessed by the DGAI score.

TABLE 1.

Examination of characteristics of participants by cumulative frequency of SSB consumption¹

	Nonconsumers (0 to <1/mo)	Occasional (1/mo to <1/wk)	Frequent (1/wk to <1/d)	Daily (≥1 d)
% (n)	28 (436)	20 (316)	44 (687)	8 (122)
Median intake, servings/wk	0.0	0.6	2.5	10.3
Age, y	67 ± 8	67 ± 9	65 ± 9	61 ± 8
Women	75 (326)	65 (206)	50 (344)	40 (49)
Current smoker	6 (26)	6 (19)	7 (47)	17 (21)
Physical activity score	35 ± 5	35 ± 5	35 ± 5	36 ± 7
BMI, kg/m2	28 ± 5	28 ± 6	28 ± 5	28 ± 6
WC, in	39 ± 6	40 ± 6	40 ± 6	41 ± 6
Lipid-lowering medication	38 (166)	41 (130)	43 (297)	36 (44)
EI	1686 ± 497	1802 ± 591	1927 ± 626	2304 ± 750
Fat, %EI	33.47 ± 7.12	33.80 ± 6.76	32.80 ± 6.10	32.47 ± 5.99
Saturated fat, %EI	11.19 ± 2.88	11.28 ± 2.74	11.05 ± 2.63	11.33 ± 2.64
Carbohydrate, %EI	45.09 ± 8.92	45.48 ± 8.52	47.15 ± 8.00	49.68 ± 8.79
Protein, %EI	19.37 ± 3.73	18.46 ± 3.24	17.42 ± 3.31	15.97 ± 3.48
Dietary fiber, g/d	19.12 ± 8.02	19.59 ± 7.93	18.83 ± 7.94	19.12 ± 8.34
Alcohol, g/d	9.03 ± 13.19	10.46 ± 14.18	11.55 ± 16.61	11.28 ± 16.46
DGAI (scale: 0–100)	64.1 ± 11.4	63.2 ± 10.9	60.7 ± 11.1	56.1 ± 11.2
SSB, servings/wk	0 ± 0.0	0.6 ± 0.2	2.9 ± 1.6	12.1 ± 5.3
Ceramide C16:0, µg/mL	0.165 ± 0.038	0.161 ± 0.034	0.166 ± 0.037	0.167 ± 0.038
Ceramide C22:0, µg/mL	0.604 ± 0.188	0.593 ± 0.155	0.630 ± 0.175	0.666 ± 0.194
Ceramide C24:0, µg/mL	2.253 ± 0.675	2.243 ± 0.563	2.324 ± 0.623	2.476 ± 0.698
Ceramide C22:0/C16:0	3.68 ± 0.83	3.71 ± 0.77	3.84 ± 0.83	4.02 ± 0.86
Ceramide C24:0/C16:0	13.79 ± 3.25	14.08 ± 3.09	14.28 ± 3.38	15.12 ± 4.02

¹ n = 1561. Values are mean ± SD or % (n). DGAI, 2010 Dietary Guidelines Adherence Index; EI, energy intake; SSB, sugar-sweetened beverage; WC, waist circumference.

Ceramide correlations

Partial Pearson correlation coefficients were adjusted for age, sex, smoking, and use of lipid-lowering medications. Correlation coefficients between concentrations of the C16:0 ceramides and the C22:0 and C24:0 ceramides were 0.64 and 0.55, respectively (both P < 0.001). Concentrations of the C22:0 and the C24:0 ceramides had the strongest correlation (r = 0.84, P < 0.001) among ceramides.

Associations between cumulative SSB consumption and circulating ceramides

Table 2 displays least-squares mean concentrations of ceramides and ceramide ratios across categories of cumulative SSB consumption. Adjusting for age, sex, smoking, use of lipid-lowering medications, total energy intake, alcohol intake, diet quality, and physical activity, concentrations of the C16:0, C22:0, and C24:0 ceramides were directly associated with frequency of SSB consumption (all P_trend < 0.05). Positive associations between concentrations of ceramides and frequency of cumulative SSB consumption remained significant after the addition of BMI to the model (all P_trend < 0.05). Daily consumers had 0.010 µg/mL higher mean concentrations of the C16:0 ceramide (95% CI: 0.002, 0.017 μg/mL) than nonconsumers (Table 3). In addition, daily consumers had 0.055 µg/mL higher mean concentrations of the C22:0 ceramide (95% CI: 0.018, 0.092 μg/mL) than nonconsumers. Each 1 SSB serving/d was associated with 0.005 μg/mL (95% CI: 0.002, 0.009 μg/mL) higher concentrations of the C16:0 ceramide and 0.028 µg/mL (95% CI: 0.010, 0.046 μg/mL) higher concentrations of the C22:0 ceramide (Table 3). Compared with nonconsumers, concentrations of the C24:0 ceramide were 0.175 µg/mL higher in daily consumers. However, when modeling cumulative SSB consumption as a continuous variable, the associations with concentrations of the C24:0 ceramide were attenuated and became nonsignificant. Finally, there were no statistically significant associations between the frequency of cumulative SSB consumption and the C22:0/C16:0 and C24:0/C16:0 ceramide ratios in either multivariable model (Table 2 and Table 3).

TABLE 2.

Adjusted least-square means of circulating ceramide concentrations and ceramide ratios by cumulative sugar-sweetened beverage category¹

	Nonconsumers (0 to <1/mo)	Occasional (1/mo to <1/wk)	Frequent (1/wk to <1/d)	Daily (≥1 d)	P trend 2	Lipid Rx Pinteraction3	Diabetes Pinteraction3
% (n)	28 (436)	20 (316)	44 (687)	8 (122)
Median intake, servings/wk	0.0	0.6	2.5	10.3
Ceramide C16:0, µg/mL
Model 1	0.161 (0.158, 0.165)	0.160 (0.156, 0.164)	0.168 (0.165, 0.170)	0.171 (0.165, 0.178)	0.002	0.018	0.18
Model 1 + BMI	0.161 (0.158, 0.165)	0.160 (0.156, 0.164)	0.168 (0.165, 0.170)	0.171 (0.165, 0.177)	0.003	0.02	0.17
Ceramide C22:0, µg/mL
Model 1	0.601 (0.585, 0.618)	0.595 (0.576, 0.614)	0.633 (0.620, 0.646)	0.656 (0.624, 0.687)	<0.001	0.47	0.06
Model 1 + BMI	0.601 (0.584, 0.617)	0.594 (0.575, 0.613)	0.634 (0.621, 0.647)	0.656 (0.624, 0.688)	<0.001	0.46	0.06
Ceramide C24:0, µg/mL
Model 1	2.251 (2.191, 2.311)	2.254 (2.186, 2.322)	2.329 (2.282, 2.375)	2.428 (2.314, 2.543)	0.004	0.56	0.015
Model 1 + BMI	2.252 (2.193, 2.312)	2.256 (2.188, 2.325)	2.327 (2.280, 2.374)	2.428 (2.313, 2.542)	0.005	0.58	0.014
Ceramide C22:0/C16:0
Model 1	3.75 (3.67, 3.82)	3.75 (3.66, 3.84)	3.81 (3.75, 3.87)	3.86 (3.71, 4.01)	0.14	—	—
Model 1 + BMI	3.74 (3.66, 3.82)	3.74 (3.66, 3.83)	3.82 (3.76, 3.88)	3.86 (3.72, 4.01)	0.11	—	—
Ceramide C24:0/C16:0
Model 1	14.07 (13.76, 14.38)	14.28 (13.92, 14.63)	14.13 (13.89, 14.37)	14.49 (13.89, 15.09)	0.33	—	—
Model 1 + BMI	14.07 (13.75, 14.38)	14.27 (13.91, 14.63)	14.13 (13.89, 14.38)	14.49 (13.89, 15.10)	0.32	—	—

¹ n = 1561. Values are least-square means and 95% CIs. Model 1 was adjusted for age, sex, smoking, lipid-lowering medication, energy intake, alcohol intake, 2010 Dietary Guidelines Adherence Index, and physical activity score. Diabetes defined as fasting plasma glucose concentrations ≥5.6 mmol/L or self-reported use of antidiabetic medications. Lipid Rx, lipid-lowering medication.

² P values for a linear test for trend across consumption categories.

³ P values for interaction term in models including the cross-product terms for dichotomous variables indicating use of Lipid Rx and diabetes status. The significance level for interactions was set at P ≤ 0.017 (0.05/3) to account for multiple tests. We found no significant interactions by BMI status (all P > 0.15, data not shown).

TABLE 3.

Associations of cumulative SSB consumption and concentrations of circulating ceramides¹

	Nonconsumers (0 to <1/mo)	Occasional (1/mo to <1/wk)2	Frequent (1/wk to <1/d)2	Daily (≥1 d)2	Per 1 serving/d3
% (n)	28 (436)	20 (316)	44 (687)	8 (122)
Ceramide C16:0, µg/mL
Model 1	Ref.	−0.001 (−0.006, 0.004)	0.006 (0.002, 0.011)	0.010 (0.002, 0.017)	0.005 (0.002, 0.009)
Model 1 + BMI	Ref.	−0.001 (−0.006, 0.004)	0.006 (0.002, 0.010)	0.010 (0.002, 0.017)	0.005 (0.002, 0.009)
Ceramide C22:0, µg/mL
Model 1	Ref.	−0.007 (−0.031, 0.018)	0.032 (0.011, 0.053)	0.054 (0.018, 0.091)	0.028 (0.010, 0.046)
Model 1 + BMI	Ref.	−0.007 (−0.032, 0.054)	0.033 (0.012, 0.054)	0.055 (0.018, 0.092)	0.028 (0.010, 0.046)
Ceramide C24:0, µg/mL
Model 1	Ref.	0.003 (−0.086, 0.093)	0.078 (0.001, 0.155)	0.178 (0.045, 0.310)	0.054 (−0.011, 0.120)
Model 1 + BMI	Ref.	0.004 (−0.086, 0.093)	0.075 (−0.002, 0.152)	0.175 (0.043, 0.308)	0.053 (−0.013, 0.118)
Ceramide C22:0/C16:0
Model 1	Ref.	0.005 (−0.110, 0.121)	0.068 (−0.031, 0.167)	0.114 (−0.057, 0.285)	0.005 (−0.007, 0.017)
Model 1 + BMI	Ref.	0.003 (−0.111, 0.117)	0.081 (−0.017, 0.179)	0.123 (−0.046, 0.292)	0.006 (−0.006, 0.018)
Ceramide C24:0/C16:0
Model 1	Ref.	0.206 (−0.264, 0.676)	0.061 (−0.343, 0.464)	0.422 (−0.274, 1.118)	−0.010 (−0.059, 0.039)
Model 1 + BMI	Ref.	0.205 (−0.265, 0.675)	0.067 (−0.337, 0.470)	0.426 (−0.270, 1.122)	−0.010 (−0.059, 0.039)

¹ n = 1561. Values are β (95% CI) unless indicated otherwise. Multivariable regression model with SSB as the independent variable and ceramides as the dependent variable. Model 1 was adjusted for age, sex, smoking, lipid-lowering medication, energy intake, alcohol intake, 2010 Dietary Guidelines Adherence Index, and physical activity score. SSB, sugar-sweetened beverage.

²β Estimates represent the difference in ceramide concentration relative to nonconsumers.

³β Estimates represent the difference in ceramide concentration per 1 SSB serving/d increase.

Effect modification in the associations between cumulative SSB consumption and circulating ceramides

In the association between cumulative SSB consumption and concentrations of the C24:0 ceramide we observed significant effect modification by diabetes status (P_interaction = 0.014; Figure 1C). In a stratified analysis, associations between cumulative SSB consumption and concentrations of the C24:0 ceramide were only statistically significant in participants with prediabetes or T2D (P_trend = 0.001; Figure 1C). Although not statistically significant, similar trends were observed in the relations of cumulative SSB consumption with concentrations of the C16:0 (P_interaction = 0.17) and C22:0 (P_interaction = 0.06) ceramides when stratifying by diabetes status (Figure 1A, B). In addition, we observed potential effect modification in the relation between cumulative SSB consumption and concentrations of the C16:0 ceramide (P_interaction = 0.02). In stratified analyses, more frequent cumulative SSB consumption was related to higher concentrations of the C16:0 ceramide only in participants using lipid-lowering medications (P_trend = 0.001; Figure 2A). There were no significant interactions by BMI status (all P  > 0.15, data not shown).

FIGURE 1.

Associations of SSB consumption and concentrations of the C16:0 ceramide (A), the C22:0 ceramide (B), and the C24:0 ceramide (C) stratified by diabetes status (n = 780 prediabetic/diabetic and n = 781 normal). Data depicted are least-square means and 95% CIs of C16:0, C22:0, and C24:0 ceramide concentrations (in micrograms per milliliter) across SSB consumption categories. Prediabetic/diabetic was defined as fasting plasma glucose concentrations ≥5.6 mmol/L or self-reported use of antidiabetic medications. Multivariable regression analysis with adjustment for age, sex, smoking, lipid-lowering medication, total energy intake, alcohol intake, 2010 Dietary Guidelines Adherence Index, physical activity score, and BMI. P values are for interaction terms (P_interaction) and for a linear test for trend across consumption categories (P_trend). SSB, sugar-sweetened beverage.

FIGURE 2.

Associations of SSB consumption and concentrations of the C16:0 ceramide (A), the C22:0 ceramide (B), and the C24:0 ceramide (C) stratified by current use of lipid-lowering medication (n = 637 users and n = 924 nonusers). Data depicted are least-square means and 95% CIs of C16:0, C22:0, and C24:0 ceramide concentrations (in micrograms per milliliter) across SSB consumption categories. Multivariable regression analysis with adjustment for age, sex, smoking, total energy intake, alcohol intake, 2010 Dietary Guidelines Adherence Index, physical activity score, and BMI. P values are for interaction terms (P_interaction) and for a linear test for trend across consumption categories (P_trend). SSB, sugar-sweetened beverage.

Secondary analyses

After additional adjustment for consumption of diet beverages, associations of SSB consumption with concentrations of all circulating ceramides and ceramide ratios were unchanged (Supplemental Table 1). Further, results were similar after multivariable analysis with the inclusion of waist circumference in place of BMI (Supplemental Table 2). In addition, results of our primary analysis were similar when including participants with prevalent CVD (Supplemental Table 3). Analyses repeated with the DGAI excluding the added sugar component were similar and slightly stronger for associations between cumulative SSB consumption and the C22:0 ceramide (Supplemental Table 4). This suggests that for some associations the complete DGAI, which includes an added sugar component, may have attenuated our results toward the null. Lastly, associations of cumulative SSB intake with circulating ceramides and ceramide ratios were similar when analyses were restricted to examining only SSB consumers (Supplemental Table 5).

Discussion

The present study found that, after adjustment for dietary, lifestyle, and adiposity variables, more frequent SSB consumption was associated with higher blood concentrations of distinct ceramides but not ceramide ratios—which is consistent with our initial hypotheses. More frequent SSB consumption was consistently associated with higher concentrations of the C16:0 and C22:0 ceramides. Associations between SSB consumption and the C24:0 ceramide were less consistent across analyses (Table 3). Further, we observed significant effect modification by only diabetes status in the association between SSB consumption and concentrations of the C24:0 ceramide. This suggests that the relation between SSB consumption and the C24:0 ceramide may be unique to individuals with poor glycemic control or reduced insulin sensitivity. Similar trends were observed in the associations of SSB consumption with concentrations of the C16:0 and C22:0 ceramides. In addition, stratified analysis particularly suggested a stronger association between SSB consumption and concentrations of the C16:0 ceramide in users of lipid-lowering medication. However, because significant effect modification was not consistent across species of ceramides, it warrants further investigation.

To our knowledge, this is the first study relating habitual SSB consumption to concentrations of circulating ceramides. SSBs in the United States are primarily sweetened with sucrose (50% glucose and 50% fructose) or high-fructose corn syrup (45% glucose and 55% fructose) (25). Previous experimental work has demonstrated that diets containing excessive amounts of fructose increase total hepatic ceramides in rodents (26–28). In nonhuman primates, the long-term feeding of a Western diet, with 20% of calories from fructose, increased concentrations of total circulating ceramides, including the C16:0, C22:0, and C24:0 species, in concurrence with reductions in insulin sensitivity (29). Although such studies do not directly translate to SSB consumption in humans, they allow for speculation that excess fructose could be a mechanism that influences circulating ceramide concentrations.

There is a lack of human studies exclusively examining the long-term associations of SSB consumption, or sucrose and/or fructose intake, with concentrations of circulating ceramides. A 3-wk over-feeding trial found that over-consumption of a high-saturated-fat diet but not a high-carbohydrate diet (excess calories from juice, SSBs, and candy) led to significant increases in concentrations of circulating C24:0 and C24:1 ceramides (9). In another trial, diets were more reflective of habitual food intake and participants assigned to a high fruit and vegetable + low refined carbohydrate diet were counseled to reduce intake of high-glycemic foods including SSBs (30). Concentrations of total and individual circulating ceramides fluctuated throughout the intervention but changes were not statistically significant after 8 wk. These trials suggest that ceramides may not be responsive to short-term alterations in dietary carbohydrate quality and/or quantity. The dysregulation of DNL by excess SSB consumption may influence ceramide concentrations by an increase in production of SFAs (e.g., palmitate) involved in ceramide synthesis, perturbations in the metabolism of triglycerides or triglyceride-rich lipoproteins, or the accumulation of visceral and ectopic fat. It is possible that such short-term dietary interventions are not sufficient to induce metabolic alterations that would have an appreciable impact on circulating ceramide concentrations. In contrast, our study suggests that habitual, consistent, and more frequent SSB consumption is associated with higher ceramide concentrations.

Prior work indicates that the relation of SSB consumption with both T2D and CVD risk is not fully explained by traditional risk factors (12, 31) and evidence suggests that dietary modification of the plasma lipidome may alter disease pathogenesis (32). Hence, distinct ceramides could in part be intermediate risk factors in the association between SSB consumption and cardiometabolic outcomes. However, it is important to note that relations between concentrations of circulating ceramides and cardiometabolic diseases differ between distinct species. Circulating concentrations of both the C16:0 and C22:0 ceramides were associated with insulin resistance after ∼5 y of follow-up in the Strong Heart Family Study (33). The C16:0 ceramide appears to be particularly detrimental to cardiometabolic health and is an independent predictor of CVD death in high-risk individuals (5). A previous analysis in the Framingham Offspring Cohort found that independent of traditional risk factors, 0.045 μg/mL higher concentrations of the C16:0 ceramide were associated with a 38% higher hazard of death from all causes in healthy participants (6). Although results of our study require replication, the observed higher concentrations of the C16:0 ceramide in daily consumers relative to nonconsumers (0.01 µg/mL) could be of clinical relevance. In contrast, associations between the C24:0 ceramide and cardiometabolic diseases are less clear. Concentrations of the C24:0 ceramide were inversely associated with mortality in the Framingham Offspring Cohort (6) but were directly associated with CVD in high-risk participants of the PREDIMED trial (34). Similarly, we observed effect modification by diabetes status in the association between SSB consumption and concentrations of the C24:0 ceramide.

Strengths of the present study include our large sample size and collection of detailed dietary, lifestyle, and clinical data in the Framingham Offspring Cohort. In addition, use of cumulative consumption reduces misclassification due to intraindividual variation in reporting and may have better captured the manifestation of metabolic intermediates in the relation between cumulative SSB consumption and concentrations of ceramides. A notable limitation is the cross-sectional design of our study. Therefore, our study is unable to discern causality between SSB consumption and ceramide concentrations. Self-reported dietary data are subject to measurement error, which may attenuate relations between SSB consumption and concentrations of circulating ceramides. In addition, the Harvard FFQ did not assess all types of SSBs (e.g., sport drinks and energy drinks). However, we expect that consumption of such beverages would be low in our sample of middle-aged and older white adults (mean age 66 y). Although the homogeneity of our sample (predominantly white, of European descent) increases the internal validity of our study, our results may not be generalizable to more racially diverse populations. The present study was limited to examining ceramide species that had previously been quantified in the Framingham Offspring Cohort. Hence, our results only pertain to the C16:0, C22:0, and C24:0 ceramides and the C22:0/C16:0 and C24:0/C16:0 ceramide ratios. Lastly, although our study adjusted for many dietary and lifestyle factors, we cannot rule out the possibility of uncontrolled or residual confounding.

In conclusion, in this cross-sectional community-based sample of middle-aged adults, we observed that more frequent cumulative SSB consumption was associated with higher concentrations of distinct ceramides. In contrast, SSB consumption was not significantly associated with ceramide ratios. These data may help to further elucidate biological mechanisms that underlie the associations between SSB consumption and higher risk of cardiometabolic diseases. Future studies should seek to examine the longitudinal relation between SSB consumption and ceramides and seek to determine if relations differ by metabolic health status.

Notes

Supported by NIH Multidisciplinary Training Program in Cardiovascular Epidemiology grant 5T32HL125232 (to RSV), NIH National Heart, Lung, and Blood Institute (NHLBI) Framingham Heart Study contracts NO1-HC-25195 and HHSN268201500001I and grants P20 HL113444 and P30 DK020579, and USDA Agricultural Research Service agreement #58-1950-4-003. RSV is supported in part by the Evans Medical Foundation and the Jay and Louis Coffman Endowment from the Department of Medicine, Boston University School of Medicine.

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the view of the NIH or the Agricultural Research Service.

Supplemental Figure 1 and Supplemental Tables 1–5 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.

Data from the Framingham Heart Study are available through the database of Genotypes and Phenotypes (in process) (1).

Abbreviations used: CVD, cardiovascular disease; DGAI, 2010 Dietary Guidelines Adherence Index; DNL, de novo lipogenesis; SSB, sugar-sweetened beverage; T2D, type 2 diabetes mellitus.