Sugar-Sweetened Beverage Consumption and Plasma Lipoprotein Cholesterol, Apolipoprotein, and Lipoprotein Particle Size Concentrations in US Adults

ABSTRACT

Background

Prospective cohort studies have found a relation between sugar-sweetened beverage (SSB) consumption (sodas and fruit drinks) and dyslipidemia. There is limited evidence linking SSB consumption to emerging features of dyslipidemia, which can be characterized by variation in lipoprotein particle size, remnant-like particle (RLP), and apolipoprotein concentrations.

Objectives

To examine the association between SSB consumption and plasma lipoprotein cholesterol, apolipoprotein, and lipoprotein particle size concentrations among US adults.

Methods

We examined participants from the Framingham Offspring Study (FOS; 1987–1995, n = 3047) and the Women's Health Study (1992, n = 26,218). Concentrations of plasma LDL cholesterol, apolipoprotein B (apoB), HDL cholesterol, apolipoprotein A1 (apoA1), triglyceride (TG), and non–HDL cholesterol, as well as total cholesterol:HDL cholesterol ratio and apoB:apoA1 ratio, were quantified in both cohorts; concentrations of apolipoprotein E, apolipoprotein C3, RLP-TG, and RLP cholesterol (RLP-C) were measured in the FOS only. Lipoprotein particle sizes were calculated from nuclear magnetic resonance signals for lipoprotein particle subclass concentrations (TG-rich lipoprotein particles [TRL-Ps]: very large, large, medium, small, and very small; LDL particles [LDL-Ps]: large, medium, and small; HDL particles [HDL-Ps]: large, medium, and small). SSB consumption was estimated from food frequency questionnaire data. We examined the associations between SSB consumption and all lipoprotein and apoprotein measures in linear regression models, adjusting for confounding factors such as lifestyle, diet, and traditional lipoprotein risk factors.

Results

SSB consumption was positively associated with LDL cholesterol, apoB, TG, RLP-TG, RLP-C, and non–HDL cholesterol concentrations and total cholesterol:HDL cholesterol and apoB:apoA1 ratios; and negatively associated with HDL cholesterol and apoA1 concentrations (P-trend range: <0.0001 to 0.008). After adjustment for traditional lipoprotein risk factors, SSB consumers had smaller LDL-P and HDL-P sizes; lower concentrations of large LDL-Ps and medium HDL-Ps; and higher concentrations of small LDL-Ps, small HDL-Ps, and large TRL-Ps (P-trend range: <0.0001 to 0.001).

Conclusions

Higher SSB consumption was associated with multiple emerging features of dyslipidemia that have been linked to higher cardiometabolic risk in US adults.

Introduction

Sugar-sweetened beverages (SSBs) are the largest single source of added dietary sugars in the United States (US) (1), contributing 24% of added sugars (2). Consumption of SSBs has been associated with increased risk for a variety of cardiometabolic disorders, including cardiovascular disease (CVD) (3–6), obesity (7), type 2 diabetes (T2D) (8), and metabolic syndrome (MetSyn) (9). A common risk factor shared by these conditions is dyslipidemia, a condition characterized by one or more of the following: elevated triglyceride (TG), elevated LDL cholesterol, elevated non–HDL cholesterol, and/or low HDL cholesterol concentrations. Although dyslipidemia encompasses a variety of lipoprotein-associated risk factors, lowering LDL cholesterol concentrations is currently the primary target for CVD risk reduction due to available and efficacious pharmacotherapy (10). The role that HDL cholesterol and TG concentrations play in the development of cardiometabolic disorders is less clear (11, 12). Although evidence from observational studies is mixed, the majority have observed adverse associations between SSB consumption and traditional dyslipidemia patterns in adults (5, 13–19).

In addition to the traditional criteria of dyslipidemia, novel lipoprotein biomarkers may reveal “distinct” dyslipidemia patterns independently associated with cardiometabolic diseases (20–22). Plasma TG and cholesterol are carried in lipoprotein particles that vary in size from small to very large, including TG-rich lipoprotein particles (TRL-Ps), LDL particles (LDL-Ps), and HDL particles (HDL-Ps) (23). Remnant-like particles (RLPs) are formed from chylomicrons and TRL-Ps during delipidation by lipoprotein lipase (24, 25). Apolipoprotein B (apoB) is associated with chylomicrons, LDL-Ps, and TRL-Ps, and apolipoprotein E (apoE) is associated with TRL-Ps, where they both can serve as a ligands for LDL receptor–mediated lipoprotein clearance. Apolipoprotein A1 (apoA1), the primary apoprotein of HDL-Ps, is a cofactor for lecithin cholesterol acyltransferase, an enzyme that catalyzes the esterification of cholesterol on HDL-Ps. Apolipoprotein C3 (apoC3), a protein present in chylomicrons and TRL-Ps, inhibits lipoprotein lipase activity, hence slowing TG clearance from plasma (10, 26, 27).

Emerging examples of distinct dyslipidemias include diabetic and atherogenic dyslipidemias. Common characteristics include high TG, small LDL-P and RLP, along with low HDL cholesterol concentrations. This pattern is commonly observed among individuals with T2D and MetSyn (20, 21, 28, 29). The utility of measuring additional lipoprotein components may be to better define individualized risk to enhance personalized clinical therapy (30–37). Animal studies suggest that consumption of fructose, a major component of SSBs, activates hepatic and enterocyte de novo lipogenesis and upregulates expression of related enzymes (38–40). Human intervention studies suggest that consumption of SSBs may promote diabetic dyslipidemia (18, 41, 42) and high carbohydrate diets may shift lipoprotein particle size profiles to atherogenic patterns (43–48). To date, no observational studies have examined the association between SSB consumption and concentrations of apolipoproteins, RLPs, and lipoprotein sizes.

The objective of the present study was to examine the association between SSB consumption and plasma lipoprotein cholesterol, apolipoprotein, and lipoprotein particle size concentrations among participants from the Framingham Offspring Study (FOS) and Women's Health Study (WHS) to generate hypotheses by which SSBs may influence distinct dyslipidemia patterns and cardiometabolic disease risk.

Methods

Subjects

The study population consisted of participants from the FOS and WHS. The Framingham Heart Study is a long-standing prospective cohort study in Framingham, Massachusetts, that began in 1948. Data from the FOS cohort (49), which includes the offspring of the original Framingham Heart Study participants, at examination 4 (1987–1991; n = 4019) and examination 5 (1991–1995; n = 3799) were used in the current study. FOS participants underwent a detailed medical history, physical examination, and standard laboratory tests during both examination periods. The WHS is a prospective cohort of North American female health care professionals aged ≥45 y (50, 51), where approximately 72% of the participants provided a voluntary blood sample (n = 28,346). The WHS began as a randomized placebo-controlled trial of low-dose aspirin, β-carotene, and vitamin E for the primary prevention of CVD and cancer and ran from 1993 to 2004 (clinicaltrials.gov: NCT00000479). The WHS continues to follow participants annually on an observational basis. Participants provided demographic, diet, lifestyle, and medical history data via standard questionnaires for both cohorts. All participants provided written informed consent before study participation. The institutional review board approved all study protocols and procedures for the current study for human research at Boston University Medical Campus, Tufts University Health Sciences, and Brigham and Women's Hospital.

A total of 3306 FOS participants provided dietary data (examination 5) and fasting plasma samples for lipoprotein measures (examinations 4 and 5). FOS participants were excluded if they provided invalid dietary data (n = 9), lipoprotein measurements were not available (examination 4: n = 182; examination 5: n = 9), or they were using lipid-lowering therapy at the time of the fasting blood draw (examination 4: n = 116; examination 5: n = 241), reducing the sample size to 2999 at examination 4 and 3047 at examination 5. A total of 28,346 participants provided baseline data in the WHS. Participants were excluded if they provided invalid dietary data or none (n = 842), lipoprotein measurements were not available (n = 395), or they were using lipid-lowering therapy (n = 891), reducing the sample size to 26,218 for this analysis. In some WHS participants, baseline blood samples were collected <8 h after a meal. Thus, results excluding these participants are presented in the Supplemental Material for the 18,805 (66%) participants who provided a baseline blood sample ≥8 h after a meal (fasting subsample).

Assessment of lipoprotein outcomes

Plasma samples from study participants in the FOS at examination 5 and the WHS at baseline were used to measure traditional lipoprotein cholesterol measures. In the FOS, HDL cholesterol, TG, and total cholesterol (TC) were measured with an Abbott Diagnostics ABA-200 analyzer (52), and LDL cholesterol was calculated according to the Friedewald equation (53) (LDL cholesterol = TC – HDL cholesterol – TG/5) and reported as not available if TG ≥ 400 mg/dL. Among WHS participants, HDL cholesterol, TG, and TC were measured using a Hitachi autoanalyzer from Roche Diagnostics, and LDL cholesterol was determined directly using a Roche homogenous assay (50). Non–HDL cholesterol concentrations were calculated as TC minus HDL cholesterol concentrations, and the TC:HDL cholesterol ratio was calculated by dividing TC by HDL cholesterol concentrations.

Plasma samples from study participants in the FOS at examination 4 (unless otherwise noted) and the WHS at baseline were used to measure emerging lipoprotein and apolipoprotein measures. Using proton nuclear magnetic resonance (NMR) spectroscopy (54), the following lipoprotein particle size concentrations were measured: TRL-P (very large, large, medium, small, and very small), LDL-P (large, medium, and small), and HDL-P (large, medium, and small). From these data, the NMR analysis software calculates TRL-P, LDL-P, and HDL-P mean size (nm diameter), which can be used to provide an assessment of lipoprotein particle size profiles. Lipoprotein particle size profiles for FOS participants at examination 4 were measured with the LPI assay version (LipoScience Inc.). This provides slightly different size groupings (Supplemental Table 1) and different units [cholesterol concentrations within each subclass (mg/dL)] than the LP4 assay version, which was used to measure lipoprotein particle size profiles for WHS participants (particle concentrations within each subclass [nmol/L for TRL-P and LDL-P; μmol/L for HDL-P]) (LP4 NMR MetaboProfile Analysis; LipoScience/LabCorp Global Research Services). The newer LP4 NMR assay provided NMR-derived concentrations of apoB and apoA1 among WHS participants, a method that is highly correlated with standard assays (r ≈ 0.95). In the FOS, apoA1 and apoB were measured using noncompetitive ELISA (Tufts University) (55, 56); apoC3 and apoE were measured at examination 5 using corresponding Wako Autokits; and RLP cholesterol (RLP-C) and RLP triglyceride (RLP-TG) concentrations were measured at examination 4 using previously described assays (57–59). In both cohorts, the apoB:apoA1 ratio was calculated by dividing apoB by apoA1.

Assessment of dietary intakes

Usual dietary intakes in the past year were estimated using the Harvard 126-item (FOS examination 5) and 131-item (WHS) semiquantitative food frequency questionnaire (FFQ) (60, 61). The FFQ consisted of a list of foods with standard serving sizes and a selection of nine frequency categories ranging from none or <1 serving per month to ≥6 servings per day. Participants with implausible energy intake were excluded from the analysis based on cohort-specific cutoffs (FOS: <600 kcal/d for men and women, ≥4000 kcal/d for women, ≥4200 kcal/d for men, and if >13 food items were left blank on the FFQ; WHS: <600 and ≥3500 kcal/d). Both FFQs have been examined for reproducibility and validity for nutrients and foods in women and men in various cohorts (60–63). Measures of apoA1, apoB, RLP, and lipoprotein particle size profiles were measured in the FOS at examination 4; however, dietary intake data were not measured at this examination period. Based on other examination periods, SSB intake does not substantially change between consecutive examinations (e.g., for SSB intake at examination 5 and 6, r² = 0.52), and estimates reflect approximate habitual SSB consumption patterns. Thus, for these lipoprotein measures measured only at examination 4, examination 5 dietary intakes were used to approximate dietary intakes at examination 4 in FOS participants.

Estimates of SSB consumption in both cohorts included the following categories 1) Coke, Pepsi, or other cola beverages with sugar; 2) caffeine-free Coke, Pepsi, or other cola beverages with sugar; 3) other carbonated beverages with sugar (e.g., 7-Up, ginger ale); and 4) Hawaiian Punch, lemonade, or other noncarbonated fruit drinks. One serving of an SSB is equivalent to 360 mL (12 fl oz). Food groupings were based on the 2020 Dietary Guidelines for Americans (2), and nutrient intakes were calculated from FFQ data by multiplying the frequency of consumption of a food item by the nutrient contents per standard serving size for the given food item.

Covariate assessment

In each cohort, participants provided general demographic, lifestyle, and medical history data via standard questionnaires. Education was assessed by asking about the highest degree or level of education that the participant had completed. Participants were grouped into categories (FOS: less than high school, graduated high school, some college, or graduated college; WHS: less than bachelor degree, bachelor degree, graduate degree). In the FOS, participants were classified as diabetic if their fasting blood glucose was ≥126 mg/dL or their nonfasting blood glucose was ≥200 mg/dL. In the WHS, participants self-reported whether they were diagnosed with diabetes. FOS participants completed a standardized physical examination, which included height and weight measurements. WHS participants provided self-reported height and weight; high correlations between self-reported and measured weights (r = 0.97) have been demonstrated in US women of similar age ( 64). BMI was calculated as weight divided by height (kg/m²). Alcohol intake was assessed by asking the number of alcoholic beverages consumed in a typical week in the previous year and expressed as g/d. Current smokers were defined as participants who reported currently smoking (WHS) or smoking regularly in the past year (FOS). Physical activity was evaluated through standard questionnaires in the FOS (examination 5 only) (65) and WHS (66). Potential confounding through other dietary components was explored through adjustment of individual dietary factors (percentage energy from saturated fat and servings per day of fruit, vegetables, whole grains, fish, and nuts/seeds).

Statistical analyses

Participants were grouped by categories of SSB consumption (servings: <1/mo, 1–4/mo, 1–2/wk, 3–7/wk, >1/d), similar to previous studies (18, 19, 67, 68). Linear (WHS) and linear mixed effects (FOS) regression models were used to examine the association between SSB consumption and LDL cholesterol, HDL cholesterol, TG, non–HDL cholesterol, TC:HDL cholesterol ratio, apoB, apoA1, apoB:apoA1, TRL-P (very large, large, medium, small, and very small), LDL-P (large, medium, and small), and HDL-P (large, medium, and small). In the FOS cohort, familial correlation was accounted for by adding a random effect in the model with a covariance structure proportional to the kinship matrix as implemented in the lmekin function of the coxme R package (https://cran.r-project.org). Similar models were used to explore the RLP-TG, RLP-C, apoE, and apoC3 outcomes that were available only in the FOS cohort. A natural logarithmic transformation was applied to TG concentrations, and quantile normalization was applied to RLP concentrations to approximate normal distributions. Due to differences in unit measures between FOS and WHS participants and skewed distributions, lipoprotein particle size concentrations (TRL-P, LDL-P, and HDL-P measures) were harmonized through quantile normalization. Four models were performed. Model 1 adjusted for age (continuous), sex (FOS only: male, female), fasting status (WHS only: ≥8 h, <8 h, or missing hours since last meal), and total energy intake (continuous). Model 2 adjusted for model 1 covariates plus education (FOS: less than high school, graduated high school, some college, graduated college; WHS: less than bachelor degree, bachelor degree, graduate degree), income (WHS only: <50,000/y, ≥$50,000/y) (69), current smoking status (yes, no), race (WHS only: non-Hispanic white, black, Asian, other), physical activity (FOS: continuous index; WHS: continuous metabolic equivalents), alcohol (g/d), current diabetes (yes, no), servings per day (vegetables, whole fruits, whole grains, nuts/seeds, and seafood), and percentage energy from saturated fat (continuous). Model 3 adjusted for model 2 covariates plus BMI (continuous), which is a marker of adiposity and could be a mediator in the association between SSB consumption and lipoprotein measures. Model 4 adjusted for model 3 covariates plus total lipoprotein concentration [TG (ln), HDL cholesterol, or LDL cholesterol for TRL-P, HDL-P, and LDL-P measures, respectively] and was applied only for lipoprotein particle size measures. Model 4 allowed us to examine the association of SSB consumption with the lipoprotein particle size measures independent of the association between SSBs and traditional lipoprotein measures. Models were run separately for the WHS and FOS cohorts, and regression coefficients and standard errors were combined through fixed effects multivariate meta-analyses (70) using the mvmeta R package. To assess for a linear trend across categories, SSB category was treated as a continuous variable, and regression coefficients and standard errors were combined through fixed effects univariate meta-analysis using the meta R package. The Cochrane Q statistic and the I² statistic were used to examine the heterogeneity of the associations between the cohorts.

Likelihood ratio testing comparing models with and without multiplicative interaction terms was used to assess effect modification by sex (male, female; FOS only) and BMI (<25, 25–29.9, ≥30). Among WHS participants, sensitivity analyses were conducted among the fasting subsample. All statistical analyses were performed using SAS (version 9.4 or higher; SAS Institute) and R (version 3.1 or higher) statistical software. All reported P values are two-sided, and results were considered statistically significant at a global P < 0.05, corrected for multiple end points using the Tukey method (71).

Results

Table 1 presents the characteristics of participants within each cohort among the highest and lowest categories of SSB consumption. Full descriptive statistics stratified by all five categories of SSB intake are displayed in Supplemental Table 2 (FOS) and Supplemental Table 3 (WHS). Among FOS participants, the highest SSB consumers were less likely to be women. In both cohorts, the highest SSB consumers were younger, more likely to smoke, and less likely to have diabetes. Additionally, WHS participants consuming more SSBs had achieved lower levels of education, were less physically active and had a higher BMI, whereas FOS participants consuming more SSBs had achieved higher levels of education and were more physically active. Lipoprotein profiles were less favorable among the highest SSB consumers. Total energy intake was higher with increasing SSB consumption, although fruit, vegetable, whole grain, seafood, and alcohol consumption was lower with increasing SSB consumption.

TABLE 1.

Characteristics of the FOS and WHS participants by cohort and category of SSB intake¹

	FOS				WHS
	Overall	<1 SSB serving/mo	>1 SSB serving/d	P-trend	Overall	<1 SSB serving/mo	>1 SSB serving/d	P-trend
n	3047	1100	312		26,218	11,751	1400
SSB servings/wk, median		0	14.0			0	14.0
Sex: women, %	54.1	64.5	34.3	<0.0001	100	100	100
Age, y	54.3 (9.8)	55.9 (9.7)	51.1 (10.2)	<0.0001	54.6 (7.0)	54.8 (7.0)	52.6 (6.0)	<0.0001
Current smoker, %	19.4	15.7	30.1	<0.0001	11.6	10.2	20.4	<0.0001
Diabetic, %	6.5	8.9	5.8	<0.0001	2.5	3.6	2.0	<0.0001
Education: bachelor degree, %	33.6	31.2	33.7	<0.0001	43.6	46.8	32.3	<0.0001
BMI, kg/m2	27.3 (5.0)	27.6 (5.3)	27.5 (4.9)	0.35	25.9 (4.9)	25.9 (4.9)	26.4 (5.7)	0.03
Physical activity index2,3	34.8 (6.1)	34.1 (5.5)	36.1 (7.3)	<0.0001	—	—	—
Physical activity,3 MET h/wk	—	—	—		5.9 (17.6)	6.9 (18.3)	3.4 (14.0)	<0.0001
Traditional plasma lipoprotein measures
LDL-C, mg/dL	126 (32)	125 (33)	124 (31)	0.44	124 (34)	123 (34)	126 (36)	<0.0001
HDL-C, mg/dL	51 (15)	53 (16)	45 (13)	<0.0001	54 (15)	55 (16)	49 (14)	<0.0001
TG,3 mg/dL	122 (88)	119 (86)	134 (106)	<0.0001	121 (90)	119 (89)	135 (111)	<0.0001
Non–HDL-C, mg/dL	153 (38)	152 (38)	154 (37)	0.01	157 (41)	156 (41)	162 (43)	<0.0001
Ratio total: HDL-C	4.4 (1.5)	4.2 (1.5)	4.8 (1.5)	<0.0001	4.2 (1.3)	4.1 (1.3)	4.6 (1.5)	<0.0001
Plasma apolipoprotein measures
apoB, mg/dL	97 (25)	97 (26)	99 (27)	0.03	94 (23)	94 (22)	96 (24)	<0.0001
apoA1, mg/dL	144 (30)	149 (33)	136 (24)	<0.0001	154 (30)	156 (30)	146 (29)	<0.0001
apoB:apoA1	0.70 (0.23)	0.67 (0.23)	0.75 (0.24)	<0.0001	0.71 (0.24)	0.69 (0.23)	0.77 (0.26)	<0.0001
apoE, mg/dL	10.1 (4.8)	10.4 (4.9)	9.9 (4.2)	0.08	—	—	—
apoC3, mg/dL	16.3 (4.5)	16.5 (4.3)	16.2 (4.7)	0.27	—	—	—
Other plasma lipoprotein measures
RLP-TG, mg/dL	18.0 (13.2)	17.6 (11.5)	21.3 (16.7)	<0.0001	—	—	—
RLP-C, mg/dL	7.1 (2.2)	7.0 (2.0)	7.6 (2.8)	0.0005	—	—	—
TRL-P4
Mean size, nm	46.2 (9.3)	45.9 (9.2)	47.9 (10.1)	<0.0001	44.1 (7.5)	44.0 (7.6)	45.2 (8)	0.0004
Very large3	3.7 (6.5)	3.5 (5.5)	4.7 (9.9)	0.0001	0.10 (0.10)	0.10 (0.10)	0.12 (0.20)	0.02
Large3	2.7 (25)	2.2 (21)	4.3 (33.9)	<0.0001	0.9 (3.9)	0.8 (3.8)	1.2 (4.7)	<0.0001
Medium3	30.6 (37)	28.4 (34.3)	34.5 (39.9)	<0.0001	12.6 (16.7)	12.0 (16.6)	14.6 (17.9)	<0.0001
Small3	14.1 (18.6)	13.6 (18.7)	14.0 (18.9)	0.79	47.1 (47.7)	45.7 (48.2)	46.6 (49.2)	<0.0001
Very small3	1.1 (6.6)	1.1 (6.8)	1.4 (5.9)	0.23	77.9 (55.1)	77.3 (54.7)	81.9 (56.8)	<0.0001
LDL-P4
Mean size, nm	20.9 (0.6)	20.9 (0.5)	20.8 (0.6)	<0.0001	20.9 (0.5)	20.9 (0.4)	20.8 (0.5)	<0.0001
Large3	53.2 (48.5)	55 (47.1)	43.5 (47.3)	<0.0001	183 (306)	196 (306)	123 (311)	<0.0001
Medium3	20.6 (32.8)	19.9 (30.4)	25.3 (29.4)	0.0005	17 (348)	17 (343)	10 (334)	0.008
Small3	13.7 (23.5)	12.5 (22.7)	17 (28.4)	<0.0001	926 (638)	913 (602)	1012 (746)	<0.0001
HDL-P4
Mean size, nm	9.2 (0.5)	9.2 (0.5)	9.0 (0.4)	<0.0001	9.0 (0.4)	9.0 (0.4)	8.9 (0.4)	<0.0001
Large	10.5 (7.1)	11.6 (7.9)	8.1 (5.6)	<0.0001	2.5 (1.5)	2.6 (1.6)	2.1 (1.4)	<0.0001
Medium	22.9 (10.2)	24.6 (10.9)	19.8 (9.1)	<0.0001	5.7 (2.8)	5.9 (2.9)	5.1 (2.6)	<0.0001
Small	17.8 (4.9)	17.3 (5.0)	18.1 (4.7)	0.003	16.4 (3.4)	16.3 (3.5)	16.8 (3.4)	<0.0001
Dietary intakes
Total energy, kcal/d	1869 (612)	1661 (541)	2368 (595)	<0.0001	1733 (529)	1612 (490)	2148 (548)	<0.0001
Saturated fat, % total energy	10.6 (2.8)	10.4 (3.0)	10.0 (2.5)	0.98	10.2 (2.5)	10.1 (2.6)	9.8 (2.4)	0.06
Servings/d
Fruits3	0.6 (1.1)	0.7 (1.1)	0.5 (1.0)	<0.0001	1.5 (1.5)	1.5 (1.4)	1.2 (1.5)	<0.0001
Vegetables3	1.7 (1.2)	1.8 (1.2)	1.6 (1.0)	0.0005	3.3 (2.7)	3.3 (2.7)	3.0 (2.7)	<0.0001
Whole grain3	0.6 (1.3)	0.7 (1.4)	0.4 (1.1)	<0.0001	1.0 (1.3)	1.0 (1.4)	0.7 (1.1)	<0.0001
Nuts/seeds3	0.2 (0.4)	0.1 (0.4)	0.2 (0.5)	<0.0001	0.1 (0.3)	0.1 (0.2)	0.1 (0.3)	0.04
Seafood3	0.3 (0.4)	0.3 (0.4)	0.2 (0.3)	0.05	0.2 (0.2)	0.2 (0.2)	0.2 (0.1)	<0.0001
Alcohol,3 g/d	1.2 (13.5)	1.1 (12.7)	0.6 (14.0)	0.14	0.3 (4.6)	0.4 (5.9)	0.1 (2.1)	<0.0001

¹Values are means (SD) unless otherwise noted. Dashes in cells indicate no data. Maximum available observations (n) for analyses. All FOS values are derived from examination 5 data except apoB, apoA1, apoB:apoA1 ratio, and all “other lipoprotein measures,” which were derived from examination 4 data. P-trend represents the P value for regression coefficients where five categories of SSB intake (serving: <1/mo, 1–4/mo, 1–2/wk, 3–7/wk, >1/d) are treated as a continuous variable in unadjusted models. Descriptive statistics for all five categories of SSB intake are displayed in Supplemental Table 2 (FOS) and Supplemental Table 3 (WHS). One serving was equivalent to the following: 360 mL, SSB; 0.5 cup, fruit; 1 cup, vegetables; 1 oz, whole grains; 1 oz, nuts/seeds; and 3.5 oz, seafood. apoA1, apolipoprotein A1; apoB, apolipoprotein B; apoC3, apolipoprotein C3; apoE, apolipoprotein E; FOS, Framingham Offspring Study; HDL-C, HDL cholesterol; HDL-P, HDL particle; LDL-C, LDL cholesterol; LDL-P, LDL particle; MET, metabolic equivalent; RLP-C, remnant-like particle cholesterol; RLP-TG, remnant-like particle triglyceride; SSB, sugar-sweetened beverage; TG, triglyceride; TRL-P, triglyceride-rich lipoprotein particle; WHS, Women's Health Study.

²Physical activity index calculated in FOS participants was previously defined (65).

³Geometric mean (IQR).

⁴In the FOS, the LPI assay measured cholesterol concentrations (mg/dL) within each lipoprotein subclass, whereas the LP4 assay used in the WHS measured particle concentrations (nmol/L for TRL-P and LDL-P; μmol/L for HDL-P) within each lipoprotein subclass.

Associations between SSB intake and lipoprotein and apolipoprotein concentrations

Table 2 presents the associations of SSB intake with lipoprotein and apolipoprotein concentrations among categories of SSB intake (reference: <1 serving per month) in FOS, WHS, and combined analyses. In the combined analyses using the fully adjusted models (model 3), participants in the highest category of SSB intake (>1 serving per day) had higher mean concentrations of LDL cholesterol (β ± SE: 2.1 ± 1.0 mg/dL), TG (0.11 ± 0.01 ln-mg/dL), non–HDL cholesterol (5.0 ± 1.2 mg/dL), TC:HDL cholesterol ratio (0.36 ± 0.04), apoB (4.6 ± 0.8), and apoB:apoA1 ratio (0.05 ± 0.007), and lower HDL cholesterol (–3.5 ± 0.4 mg/dL) and apoA1 (−4.1 ± 0.7 mg/dL) concentrations, compared with those in the lowest category of SSB intake (<1 serving per month). Trend analyses indicated a statistically significant linear trend across the five categories of SSB intake for all lipoprotein and apolipoprotein measures (P-trend <0.001). In analyses restricted to the FOS cohort, participants in the highest category of SSB intake (>1 serving per day) had higher mean concentrations of RLP-TG (β ± SE: 0.22 ± 0.08 mg/dL; P-trend = 0.001) and RLP-C (0.13 ± 0.08 mg/dL; P-trend = 0.008). No significant associations were observed for apoE or apoC3 concentrations in the FOS cohort. No significant interactions (P < 0.05) between SSB consumption and sex or BMI were observed in fully adjusted models. We observed similar results when analyses were limited to the fasting subsample of WHS participants (Supplemental Table 4).

TABLE 2.

Associations between SSB intake and plasma lipoprotein and apolipoprotein concentrations in the FOS and WHS

	SSB serving,1 β (SE)
	<1/mo	1–4/mo	1–2/wk	3–7/wk	>1/d	P-trend	Adjusted P-trend2	P-heterogeneity
FOS3
n	1100	710	289	635	312
SSB servings/wk,1 median	0	0.5	1.5	4.0	14.0
LDL-C, mg/dL	Ref	1.4 (1.6)	3.0 (2.2)	3.0 (1.7)	0.9 (2.5)	0.18	0.48
HDL-C, mg/dL	Ref	−0.6 (0.6)	−1.3 (0.8)	−2.3 (0.7)	−1.6 (0.9)	0.0009	0.002
TG, ln-mg/dL	Ref	0.04 (0.02)	0.04 (0.03)	0.09 (0.03)	0.13 (0.04)	0.0001	0.0001
Non–HDL-C, mg/dL	Ref	2.5 (1.7)	3.8 (2.4)	5.5 (1.9)	4.2 (2.7)	0.008	0.02
TC:HDL-C	Ref	0.02 (0.01)	0.05 (0.01)	0.04 (0.01)	0.04 (0.02)	0.0001	0.0001
RLP-TG, mg/dL	Ref	−0.01 (0.05)	0.15 (0.07)	0.15 (0.06)	0.22 (0.08)	0.001	0.001
RLP-C, mg/dL	Ref	−0.001 (0.05)	0.2 (0.07)	0.14 (0.05)	0.13 (0.08)	0.003	0.008
apoB, mg/dL	Ref	1.1 (1.1)	6.3 (1.6)	2.6 (1.2)	4.1 (1.8)	0.003	0.008
apoA1, mg/dL	Ref	−2.3 (1.3)	−1.7 (1.8)	−4.1 (1.4)	−2.4 (2.1)	0.02	0.06
apoB:apoA1	Ref	0.07 (0.06)	0.14 (0.09)	0.26 (0.07)	0.23 (0.10)	0.0003	0.0005
apoE, mg/dL	Ref	−0.4 (0.3)	−0.2 (0.4)	0.1 (0.3)	0.2 (0.4)	0.54	0.93
apoC3, mg/dL	Ref	0.2 (0.2)	0.3 (0.3)	0.4 (0.2)	0.5 (0.3)	0.07	0.20
WHS3
n	11,751	5724	3384	3959	1400
SSB servings/wk,1 median	0	0.5	1.4	4.0	14.0
LDL-C, mg/dL	Ref	0.4 (0.6)	1.8 (0.7)	1.9 (0.7)	2.3 (1.1)	0.0005	0.001
HDL-C, mg/dL	Ref	−1.0 (0.2)	−2.1 (0.3)	−2.4 (0.3)	−3.9 (0.4)	<0.0001	<0.0001
TG, ln-mg/dL	Ref	0.02 (0.01)	0.05 (0.01)	0.07 (0.01)	0.10 (0.02)	<0.0001	<0.0001
Non–HDL-C, mg/dL	Ref	0.6 (0.7)	2.6 (0.8)	3.6 (0.8)	5.2 (1.3)	<0.0001	<0.0001
TC:HDL-C	Ref	0.07 (0.02)	0.17 (0.02)	0.21 (0.02)	0.37 (0.04)	<0.0001	<0.0001
apoB, mg/dL	Ref	0.8 (0.4)	2.3 (0.5)	3.2 (0.5)	4.7 (0.8)	<0.0001	<0.0001
apoA1, mg/dL	Ref	−1.6 (0.4)	−3.0 (0.5)	−2.3 (0.5)	−4.3 (0.8)	<0.0001	<0.0001
apoB:apoA1	Ref	0.01 (0.004)	0.03 (0.005)	0.04 (0.004)	0.06 (0.01)	<0.0001	<0.0001
Combined results4
LDL-C, mg/dL	Ref	0.5 (0.5)	1.9 (0.7)	2.0 (0.6)	2.1 (1.0)	0.0002	0.0006	0.94
HDL-C, mg/dL	Ref	−1.0 (0.2)	−2.0. (0.3)	−2.4 (0.3)	−3.5 (0.4)	<0.0001	<0.0001	0.19
TG, ln-mg/dL	Ref	0.02 (0.01)	0.05 (0.01)	0.07 (0.01)	0.11 (0.01)	<0.0001	<0.0001	0.83
Non–HDL-C, mg/dL	Ref	0.8 (0.6)	2.7 (0.8)	3.8 (0.7)	5.0 (1.2)	<0.0001	<0.0001	0.86
TC:HDL-C	Ref	0.07 (0.02)	0.17 (0.02)	0.22 (0.02)	0.36 (0.04)	<0.0001	<0.0001	0.55
apoB, mg/dL	Ref	0.8 (0.4)	2.8 (0.5)	3.1 (0.5)	4.6 (0.8)	<0.0001	<0.0001	0.07
apoA1, mg/dL	Ref	−1.6 (0.39)	−2.9 (0.5)	−2.6 (0.5)	−4.1 (0.7)	<0.0001	<0.0001	0.41
apoB:apoA1	Ref	0.01 (0.004)	0.03 (0.004)	0.04 (0.004)	0.05 (0.007)	<0.0001	<0.0001	0.36

¹One serving of SSB is equivalent to 360 mL. apoA1, apolipoprotein A1; apoB, apolipoprotein B; apoC3, apolipoprotein C3; apoE, apolipoprotein E; FOS, Framingham Offspring Study; HDL-C, HDL cholesterol; LDL-C, LDL cholesterol; Ref, reference; RLP-C, remnant-like particle cholesterol; RLP-TG, remnant-like particle triglyceride; SSB, sugar-sweetened beverage; TC, total cholesterol; TG, triglyceride; WHS, Women's Health Study.

²Adjusted for multiple end points using the Tukey method (94).

³Values are regression coefficients for SSB intake in mixed effects models accounting for family structure (FOS) and generalized linear models (WHS) adjusted for the following: age, sex (FOS only), fasting status (WHS only), total energy intake, smoking status, education status, income (WHS only), current diabetes status, physical activity, alcohol intake, body mass index, whole fruit intake, vegetable intake, whole grains intake, seafood intake, nuts/seeds intake, and saturated fatty acid intake (percentage of total energy).

⁴Study estimates from the two cohorts were combined through fixed effects meta-analyses.

Associations between SSB intake and lipoprotein particle size measures

Table 3 and Figure 1 present the mean differences in lipoprotein particle size concentrations between the highest SSB consumers (>1 serving per day) and the lowest (<1 serving per month), with the P-trend across five categories of SSB intake in FOS, WHS, and combined analyses. After adjustment for traditional lipoprotein measures and potential confounding factors (model 4), the highest SSB consumers compared with the lowest had a smaller mean LDL-P size (β ± SE: −0.10 ± 0.01 nm; P-trend <0.0001) and HDL-P size (−0.02 ± 0.01 nm; P-trend = 0.001), with lower concentrations of large LDL-Ps (−0.18 ± 0.03; P-trend <0.0001) and medium HDL-Ps (−0.10 ± 0.03; P-trend <0.0001) and higher concentrations of small LDL-Ps (0.14 ± 0.03; P-trend <0.0001) and small HDL-Ps (0.16 ± 0.03; P-trend <0.0001). Among TRL-P measures, SSB was significantly associated only with higher concentrations of large TRL-Ps (β ± SE: 0.06 ± 0.02; P-trend = 0.0009). There was no evidence of heterogeneity between the cohorts for these significant associations (P > 0.05). However, significant heterogeneity was observed for medium LDL-Ps (P = 0.0002), where higher SSB consumption was associated with higher concentrations of medium LDL-Ps only among FOS participants (P-trend = 0.001). A similar trend was not observed among WHS participants (P-trend = 0.16), which is likely due to differences in the way that small/medium particles were grouped in the two versions of the LipoProfile assay (Supplemental Table 1).

TABLE 3.

Associations between SSB intake and plasma lipoprotein particle size concentrations in the FOS and WHS¹

	FOS (n = 3047)		WHS (n = 26,218)		Combined results2
	β (SE)3	P-trend	β (SE)3	P-trend	β (SE)3	P-trend	Adjusted P-trend4	P-heterogeneity
TRL-P
Mean size, nm	0.57 (0.56)	0.39	0.33 (0.19)	0.51	0.36 (0.18)	0.38	0.80	0.41
Very Large	0.06 (0.06)	0.28	0.02 (0.03)	0.58	0.03 (0.03)	0.90	0.99	0.75
Large	0.06 (0.05)	0.21	0.06 (0.02)	0.0008	0.06 (0.02)	0.0003	0.0009	0.99
Medium	−0.03 (0.05)	0.55	0.001 (0.02)	0.70	−0.004 (0.02)	0.89	0.99	0.76
Small	0.04 (0.08)	0.35	−0.02 (0.03)	0.23	−0.02 (0.03)	0.15	0.37	0.46
Very small	0.01 (0.08)	0.62	0.01 (0.03)	0.38	0.01 (0.03)	0.32	0.71	0.85
LDL-P
Mean size, nm	−0.08 (0.04)	0.02	−0.10 (0.01)	<0.0001	−0.10 (0.01)	<0.0001	<0.0001	0.86
Large	0.08 (0.07)	0.23	−0.23 (0.03)	<0.0001	−0.18 (0.03)	<0.0001	<0.0001	0.72
Medium	0.22 (0.07)	0.001	−0.06 (0.03)	0.16	−0.02 (0.03)	0.86	0.99	0.0002
Small	−0.17 (0.08)	0.01	0.19 (0.03)	<0.0001	0.14 (0.03)	<0.0001	<0.0001	0.57
HDL-P
Mean size, nm	−0.06 (0.02)	0.05	−0.01 (0.01)	0.002	−0.02 (0.01)	0.0004	0.001	0.45
Large	−0.10 (0.04)	0.31	−0.02 (0.02)	0.28	−0.04 (0.02)	0.16	0.41	0.53
Medium	−0.03 (0.05)	0.61	−0.12 (0.03)	<0.0001	−0.10 (0.03)	<0.0001	<0.0001	0.11
Small	0.14 (0.08)	0.11	0.16 (0.03)	<0.0001	0.16 (0.03)	<0.0001	<0.0001	0.08

¹Values are regression coefficients for the highest category of SSB intake [serving: >1/d; n = 1400 (WHS: median = 14.0/wk), n = 312 (FHS: median = 14.0/wk)] compared with the lowest category of SSB intake [serving: <1/mo; n = 11,751 (WHS: median = 0.0/d), n = 1100 (FHS: median = 0.0/d)] on quantile-normalized particle concentrations using mixed effects models accounting for family structure in FHS and generalized linear models in WHS. Models were adjusted for the following: age, sex (FHS only), fasting status (WHS only), total energy intake, smoking status, education status, income (WHS only), current diabetes status, physical activity, alcohol intake, body mass index, whole fruit intake, vegetable intake, whole grains intake, seafood intake, nuts/seeds intake, saturated fatty acid intake (percentage of total energy), and total lipid measure (triglyceride, LDL cholesterol, and HDL cholesterol concentrations for TRL-P, LDL-P, and HDL-P concentrations, respectively). P-trend represents the P value for regression coefficients where the category of SSB intake (servings: <1/mo, 1–4/mo, 1–2/wk, 3–7/wk, >1/d) is treated as a continuous variable. One serving of SSB is equivalent to 360 mL. FOS, Framingham Offspring Study; HDL-P, HDL particle; LDL-P, LDL particle; SSB, sugar-sweetened beverage; TRL-P, triglyceride-rich lipoprotein particle; WHS, Women's Health Study.

²Study estimates from the two cohorts were combined through fixed effects meta-analyses.

³SSB servings: >1/d vs. <1/mo.

⁴Adjusted for multiple end points using the Tukey method (94).

FIGURE 1.

Mean difference in plasma lipoprotein particle size concentrations among individuals consuming >1 SSB serving per day compared with <1 SSB serving per month in a fixed effect meta-analysis of the Framingham Offspring Study (FOS; n = 3047) and Women's Health Study (WHS; n = 26,218). Values are regression coefficients (β) comparing the highest SSB consumers (>1 SSB serving per day) with the lowest (<1 SSB serving per month) in fixed effect meta-analyses on quantile-normalized particle concentrations. Meta-analyses are based on mixed effects models accounting for family structure in FOS and generalized linear models in WHS, adjusted for the following. Model 1: age, sex (FOS only), fasting status (WHS only), and total energy intake. Model 2: model 1 covariates plus smoking status, education status, income (WHS only), current diabetes status, physical activity, alcohol intake, whole fruit intake, vegetable intake, whole grains intake, seafood intake, nuts/seeds intake, and saturated fatty acid intake (percentage of total energy). Model 3: model 2 covariates plus body mass index. Model 4: model 3 covariates plus total lipid measure (triglyceride, LDL cholesterol, and HDL cholesterol concentrations for TRL-P, LDL-P, and HDL-P concentrations, respectively). One SSB serving is equivalent to 360 mL. Horizontal bars indicate 95% CIs. HDL-P, HDL particle; LDL-P, LDL particle; SSB, sugar-sweetened beverage; TRL-P, triglyceride-rich particle.

Associations between all categories of SSB consumption and lipoprotein particle sizes from the multivariate meta-analysis in all four covariate models are presented in Figure 1 and Supplemental Table 5. Before adjusting for total TG concentrations (model 3), higher SSB consumption (trend across categories: <1 serving per month to >1 serving per day) was significantly associated with larger TRL-P particle size [β ± SE (serving: >1/d compared with <1/mo): 1.21 ± 0.22 nm; P-trend <0.0001), higher concentrations of all TRL-P particle fractions (very large: 0.10 ± 0.03; P-trend = 0.0009; large: 0.20 ± 0.03; P-trend <0.0001; medium: 0.14 ± 0.03; P-trend <0.0001; small: 0.01 ± 0.03; P-trend = 0.05: very small: 0.09 ± 0.03; P-trend <0.0001), and lower concentrations of large HDL-Ps (–0.20 ± 0.03; P-trend <0.0001). All other associations were similar across all covariate models. Similar associations between SSB consumption and lipoprotein particle sizes were observed when analyses were limited to the fasting subsample of WHS participants (Supplemental Table 4), and no significant interactions (P < 0.05) were observed by sex or BMI.

Discussion

In this study of two large US cohorts, SSB consumption was significantly associated with concentrations of lipoprotein cholesterol, apolipoproteins, and lipoprotein particle size concentrations that have been linked to adverse cardiometabolic outcomes. We identified novel associations between SSB intake and higher apoB:apoA1 ratio and concentrations of non–HDL cholesterol, apoB, RLP-TG and RLP-C; lower concentrations of apoA1; and smaller LDL-P and HDL-P size. We also replicated previously observed associations of SSB consumption with TC:HDL cholesterol ratio and HDL cholesterol, LDL cholesterol, and TG concentrations.

Our analysis is the largest to date that examines the associations between SSB intake and a wide range of lipoprotein concentrations. These results are consistent with several other studies observing that higher SSB consumption associates with lower concentrations of HDL cholesterol and higher concentrations of LDL cholesterol, TG, and TC:HDL cholesterol ratio (5, 14, 17–19, 68, 72). In addition to previously observed associations, our results indicate that SSB intake is associated with non–HDL cholesterol, apoB, apoA1, RLP-C, and RLP-TG concentrations and apoB:apoA1 ratio, all in the direction that has been associated with increased cardiometabolic risk in previous studies (10, 24, 30–36). Although measurement of apoB and apoA1 concentrations are not currently recommended over lipoprotein measures in clinical settings (36, 73), emerging data suggest that these measures can improve risk assessment and their elevation may confer greater risk over traditionally measured clinical risk factors (10, 26, 74). RLP concentrations are another emerging CVD risk factor (25, 75–77), and to date, few interventions have considered whether dietary sugars or SSBs influence RLP concentrations. One randomized controlled trial with 48 participants observed that consumption of SSBs at 25% of total energy intake over 2 weeks resulted in significant increases in RLP-C and RLP-TG concentrations (41). Thus, the findings suggest that reducing SSB consumption may result in concomitant RLP reduction.

The results of our study indicate that SSB consumption is significantly associated with smaller HDL-P and LDL-P size, independent of total HDL cholesterol and LDL cholesterol concentrations. Observations included a significant negative association between SSB consumption and large LDL-P and medium HDL-P concentrations and a positive association between SSB consumption and small LDL-P, small HDL-P, and large TRL-P concentrations. Despite the controversy around the clinical utility of these measures of lipoprotein particle size (78), in observational studies, smaller HDL-P and LDL-P size and larger TRL-P size have been associated with higher risk for T2D (34, 37, 79, 80), hypertension (22), MetSyn (52), peripheral artery disease (74), and CVD (74, 81–83), independent of traditional lipoprotein measures. Thus, continued investigation of the function and determinants of lipoprotein particle size is warranted to understand its unique contribution in the atherosclerotic process (78, 84). For example, in vitro studies have demonstrated that smaller LDL-Ps have an increased affinity for LDL binding sites, which may increase cellular uptake rates in arterial tissue (85, 86). Further investigation of the mechanisms by which lipoprotein particle size and type may influence cardiometabolic risk is required (i.e., through differences in cholesterol and TG content by particle size, particle concentration, or functional properties related to particle size). These investigations may lead to novel biomarker discovery that could improve cardiometabolic disease risk prediction and extend insights gained from the current study related to high SSB consumption and dyslipidemia patterns.

The findings in this study are novel as no previous observational studies have investigated the association between SSB consumption and lipoprotein particle size concentrations among adults. One cross-sectional study among 74 Swiss children observed that high total fructose intake was the only significant dietary predictor of LDL-P size out of 11 categories of macronutrient intake (87). Two small intervention studies (<50 participants) comparing consumption of glucose-, fructose-, and sugar-sweetened beverages within ranges of normal intake over 2- to 3-wk periods also observed that consumption of either fructose- or sugar-sweetened beverages led to lower LDL-P size (41, 42). Several randomized controlled trials have examined the effect of low-carbohydrate diets on LDL-P size; a meta-analysis of 16 studies indicated that carbohydrate restriction reduces the number of small LDL-Ps (48). Our findings are consistent with these prior studies, where we observed higher concentrations of small LDL-Ps among the highest SSB consumers.

Investigators have also observed significant associations between poor overall diet quality and lower plasma concentrations of large HDL-Ps (88) and higher concentrations of small HDL-Ps and medium and large TRL-Ps (89). These associations are in the same direction as the associations of SSB consumption with smaller HDL-P and larger TRL-P size that we observed in our study, despite the attenuation of the association with TRL-P size after adjustment for TG concentrations. Large prospective studies have observed that greater concentrations of large TRL-Ps are associated with a higher risk for T2D compared with small TRL-Ps (34, 79) and that concentrations of large TRL-Ps were elevated among adults with nonalcoholic fatty liver disease (90). Interestingly, two small intervention studies that compared low-carbohydrate diets and low-fat weight loss diets for 3–6 mo observed larger decreases in concentrations of large TRL-Ps compared with small TRL-Ps (91, 92) for the low-carbohydrate diet, suggesting a potential role of diets low in carbohydrates on TRL-P size. Thus, the combination of data from these previous studies and this current study suggests that further research into differences in lipoprotein metabolism could reveal new mechanisms by which SSB consumption may increase risk for cardiometabolic diseases, particularly diabetic dyslipidemia. Overall, the effect sizes of the observed associations are relatively small, and larger studies with longitudinal follow-up are needed to determine the potential clinical relevance of these findings.

The strengths of our study include a large sample size, the ability to adjust for multiple confounding factors, and the ability to examine a wide range of lipoprotein concentrations in two independent cohorts. However, the research design also has some limitations. The design of this study limits our ability to infer causality between SSB consumption and the outcomes of interest, and the use of self-reported dietary data can lead to misclassification of food and nutrient intakes. These FFQs did not include an exhaustive list of all potential sources of SSBs, such as consumption of sweetened coffee/tea. However, estimates of beverage consumption during the time of data collection (1987–1995) suggest that consumption of coffee/tea (sweetening not captured) was low compared with consumption of SSBs (93). Among FOS participants, measurements of apoA1, apoB, RLP, and lipoprotein particle size concentrations were derived from blood draws at examination 4, whereas dietary intakes were estimated at examination 5. Thus, a limitation in the FOS analysis is that estimated dietary intakes may not reflect dietary intakes at the time of the blood draw. However, the consistency of the associations between SSB consumption and lipoprotein particle size concentrations in the FOS and WHS adds confidence that our estimates of dietary intakes are informative. Generalizability of our study is also limited by our sample of predominantly European-descent adults and by the inclusion of only women and health professionals in the WHS; hence, these participants likely have a higher socioeconomic status than the general population.

In conclusion, our findings suggest that higher consumption of SSBs is associated with multiple measures of plasma lipoprotein concentrations that have been linked to adverse cardiometabolic outcomes, including traditional and emerging measures of lipoprotein cholesterol, apolipoprotein, and lipoprotein particle concentrations. These data suggest that differences in lipoprotein particle sizes are a potential pathway by which SSB intake may increase risk for cardiometabolic diseases.

Notes

This work is supported by NIH 5T32HL069772-15 (DEH), NIH 2T32CA009001-39 (DEH), AHA 16CSA28590003 (MAH, NMM, DEH), NIH R01DK100425 (MAH), NIH R01DK121710 (MAH), K08 HL112845 (CES), and USDA Agricultural Research Service agreements 58-1950-4-003 (NMM, PFJ) and 588-1950-9-001 (AHL). The Women's Health Study work was supported by the NIH (CA-047988, HL-043851, HL-080467, HL-099355, and UM1CA182913 to Women's Health Study; R01 HL134811, HL 117861, and K24 HL136852 to SM), with additional funding from the Molito Family Trust and the National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK112940 to SM). From the Framingham Heart Study of the National Heart, Lung, and Blood Institute of the NIH and Boston University School of Medicine: this project has been funded in whole or in part with federal funds from the National Heart, Lung, and Blood Institute, NIH, Department of Health and Human Services, under contract 75N92019D00031.

Author disclosures: SM has received institutional research grant support from Atherotech Diagnostics for research outside the current work, has received personal fees for scientific advisory board participation for Pfizer and Quest Diagnostics outside the current work, and is listed as coinventor on a patent for a nuclear magnetic resonance–measured biomarker (GlycA) with colorectal cancer risk. Nuclear magnetic resonance measurements were provided by LipoScience (LabCorp), Inc., which had no role in the design and conduct of this study or the interpretation of the current study data or decision to submit for publication. All other authors report no conflicts of interest.

Supplemental Material 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/jn/.

Abbreviations used: apoA1, apolipoprotein A1; apoB, apolipoprotein B; apoC3, apolipoprotein C3; apoE, apolipoprotein E; CVD, cardiovascular disease; FOS, Framingham Offspring Study; HDL-C, HDL cholesterol; HDL-P, HDL particle; LDL-C, LDL cholesterol; LDL-P, LDL particle; MetSyn, metabolic syndrome; NMR, nuclear magnetic resonance; RLP, remnant-like particle; RLP-C, remnant-like particle cholesterol; RLP-TG, remnant-like particle triglyceride; SSB, sugar-sweetened beverage; TC, total cholesterol; TG, triglyceride; TRL-P, triglyceride-rich lipoprotein particle; T2D, type 2 diabetes; WHS, Women's Health Study.

Data Availability

The Framingham Heart Study data sets used for the described analyses were obtained from dbGaP at http://www.ncbi.nlm.nih.gov/sites/entrez?db=gap through accession phs000007.v29.p10.