Associations of dietary sugar types with coronary heart disease risk: a prospective cohort study

Abstract

Background

Higher intake of total sugar has been linked with coronary heart disease (CHD) risk, but the role of individual sugars, particularly fructose, is uncertain.

Objectives

This study aimed to investigate the associations of individual dietary sugars with CHD risk.

Methods

In prospective cohort studies, we followed 76,815 women (Nurses’ Health Study, 1980-2020) and 38,878 men (Health Professionals Follow-up Study, 1986-2016). Sugar and carbohydrate intake, including total fructose equivalents ([TFE] from fructose monosaccharides and sucrose), total glucose equivalents ([TGE] from glucose monosaccharides, disaccharides, and starch), and other sugar types, was measured every 2 to 4 y by semiquantitative food frequency questionnaires.

Results

We documented 9,723 incident CHD cases over 40 years. In isocaloric substitution models with total fat as a comparison nutrient, comparing extreme quintiles of intake, hazard ratios (HRs), 95% confidence interval [CI]) for CHD risk were 1.31 (1.20 to 1.42; P_trend < 0.001) for TGE and 1.03 (0.94 to 1.11; P_trend = 0.25) for TFE. TFE from fruits and vegetables was not associated with CHD risk (P_trend = 0.70), but TFE from added sugar and juice was associated with CHD risk (HR: 1.12, 95% CI: 1.04 to 1.20; P_trend < 0.01). Intakes of total sugars and added sugar were positively associated with CHD risk (HRs: 1.16, 95% CI: 1.07 to 1.26, P_trend < 0.001; 1.08, 95% CI: 0.99 to 1.16, P_trend = 0.04).

Conclusions

Intakes of TGE, total sugar, added sugar, and fructose from added sugar and juice were associated with higher CHD risk, but TFE and fructose from fruits and vegetables were not.

Introduction

The role of dietary carbohydrates in cardiovascular health has been a subject of particular interest. Previous studies have suggested that types of carbohydrate intake were more important than total carbohydrate intake for determining coronary heart disease (CHD) risk [1, 2]. Although evidence for higher risks of CHD with some carbohydrate sources like sugar-sweetened beverages (SSB) is clear [[3], [4], [5], [6]], findings on the impact of individual sugars on incident CHD and CHD risk factors have been mixed and limited [[7], [8], [9]]. Typical Western diets are composed of many foods with large amounts of starch and sugar, which raise blood glucose more than some foods naturally high in sugar, like fruit [10]. Individual types of sugar in the diet, such as glucose or fructose, may also have differential impacts on CHD risk [11].

Some carbohydrate polysaccharides and disaccharides, such as starch, sucrose, and lactose, are rapidly metabolized into monosaccharides, predominantly fructose and glucose. After absorption into the blood, individual monosaccharides are subject to different metabolic pathways. Fructose bypasses regulatory steps in glycolysis, which can contribute to excess de novo lipogenesis and increased production and secretion of very low-density lipoprotein from the liver [12, 13]. High intakes of fructose can lead to increased postprandial triglycerides and LDL cholesterol, whereas these effects are not seen with equivalent amounts of glucose [14, 15]. Some investigators have, therefore, hypothesized that fructose intake would be especially harmful [16]. However, prospective cohort studies investigating the specific role of fructose on CHD incidence have been limited. Therefore, we investigated the associations of types of dietary sugars, particularly fructose and glucose, with CHD risk in two prospective cohort studies. Here, we define dietary sugars to include carbohydrate sources rapidly digested into monosaccharides (e.g., starch due to its rapid breakdown into glucose). We considered the impact of total glucose and total fructose from all sources in the diet on incident CHD. We hypothesized that total glucose equivalents (TGE), total fructose equivalents (TFE), and added sugar intakes are positively associated with CHD risk, but intake of sugar intrinsic to foods (e.g., sugars from fruit and 100% fruit juices, but not added sugar) is not associated with higher risk.

Methods

Study population

The Nurses’ Health Study (NHS) is a prospective cohort study that began in 1976 (recruitment June to December 1976), following 121,700 female nurses from 30 to 55 y of age. In 1980, 92,468 participants completed a baseline semiquantitative food frequency questionnaire (SFFQ). The Health Professionals Follow-up Study (HPFS) is a prospective cohort study that began in 1986 (recruitment from January to June 1986), following 51,529 male health professionals from 40 to 75 y of age. HPFS participants also completed a baseline SFFQ in 1986. Every 2 to 4 y, participants in both cohorts provided information on diet and lifestyle factors, medical history, and newly diagnosed diseases through mailed questionnaires. The study protocol was approved by the institutional review boards of the Harvard T.H. Chan School of Public Health and Brigham and Women’s Hospital (IRB approval number: 1999P0011114). Voluntarily returning the self-administered questionnaire was considered informed consent in both cohorts.

For the current study, we excluded participants who had a history of diabetes, cardiovascular disease, and cancer. Participants must have completed an SFFQ and not reported implausible energy intakes [17] (total energy intake <600 or >3500 kcal/d for women and <800 or >4200 kcal/d for men) at baseline (1980 in NHS; 1986 in HPFS). The final sample size was 76,815 women and 38,878 men (Supplemental Figure 1).

Dietary assessment

Diet was assessed with SFFQ every 2 to 4 y for a total of 9 SFFQs in NHS through 2010 and 8 SFFQs in HPFS through 2014. In each SFFQ, participants reported how often a specified portion size of each food was consumed on average during the previous year. The 1980 SFFQ included 61 items; this was expanded in 1984 to 126 items, and subsequent SFFQs were similar, with a few questions varying from year to year. We collected detailed information about food and beverage types, including if items were sweetened by sugar or artificial sweeteners. We calculated mean daily nutrient and total energy intakes by multiplying the frequency of consumption by nutrient content of the specified portion size of each food. Nutrient contents were based on the Harvard University Food Composition Table [18]. We calculated TFE by adding fructose monosaccharides and fructose from sucrose and TGE by adding glucose monosaccharides and glucose from sucrose, maltose, lactose, and starch. As a large proportion of dietary starch in the US population comes from refined grains, which are rapidly digested into glucose, starch was considered a component of TGE [19, 20]. Additionally, we considered TFE from whole fruits and vegetables as well as TFE from added sugar and fruit juice. The main exposure variables were monosaccharide sugar equivalents, including TFE and TGE, as well as other related carbohydrate types, including total sugar, added sugar, sugar intrinsic to foods (e.g., sugar in foods such as fruit, 100% fruit juice, and milk), sucrose, lactose, and starch. Total sugar includes added sugar and sugars intrinsic to foods. Total carbohydrate and total sugar intake measured by SFFQs have been validated by comparison with that measured by two 7-d dietary records (7DDR) administered 6 mo apart with correlations for total carbohydrates of r = 0.80 in HPFS and r = 0.74 in NHS and correlations for total sugar of r = 0.77 in HPFS and r = 0.75 in NHS between SFFQ and 7DDR [21, 22].

Ascertainment of CHD

Our primary endpoint of CHD included nonfatal myocardial infarction and CHD death, which were identified through medical record review and death certificates [23]. Acute myocardial infarction (MI) was confirmed if the World Health Organization criteria for MI were met; these require symptoms, electrocardiographic changes, or elevated cardiac enzymes [24]. Additionally, MIs requiring hospital admission that were also documented by phone interview or letter were considered probable. We included all confirmed and probable cases as results have been similar in previous analyses [1].

Covariates

Study participants completed biennial questionnaires on which they updated information on weight and medical history, including menopausal status, smoking, and diagnosis of hypertension or hypercholesterolemia. Alcohol intake was reported every 2 to 4 y and was characterized as grams per day. Physical activity was characterized as a metabolic equivalent task (MET) hours per week. BMI was calculated as kilograms divided by height in meters squared. Self-reported covariate measures, including alcohol intake, physical activity, and weight, have been previously validated [[25], [26], [27], [28], [29]].

Statistical analyses

Participants were followed from the initial questionnaire return (1980 in NHS; 1986 in HPFS) through 2020 in NHS and 2016 in HPFS, date of initial confirmed CHD event, death, or loss to follow-up, whichever occurred earliest. We calculated a correlation matrix among different types of sugar and fiber, starch, and glycemic load assessed the middle of the follow-up (1998 for both cohorts) using Pearson correlation coefficients. Additionally, to provide context for understanding the overall dietary contributions of carbohydrates and TFE, we calculated the average energy intake of carbohydrate sources and TFE sources at the middle of the follow-up using a weighted mean. All macronutrient variables were modeled as percent contribution of total calories, which accounted for any differences in specific questions among SFFQs during follow-up. We calculated cumulative average intakes of percentages of energy intake from total sugar, added sugar, and specific types of sugars, carbohydrates, and other nutrients by averaging all repeated SFFQ data up to the beginning of each 2-y follow-up interval as this is most representative of long-term diet and dampens within-person variation. As diet was measured repeatedly over time and calculated as cumulative average intakes, no participant included in the analysis had missing dietary information.

Participants were categorized into quintiles according to population distributions of sugar intakes. Cox proportional hazard models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CI) between intake of sugar types and CHD risk in each cohort by comparing each quintile of intake with the lowest quintile. To determine a linear trend, we assigned the median intake level to each quintile, modeled this variable continuously, and used the Wald test to determine statistical significance. We built isocaloric substitution models that included total energy intake, percent energy from dietary protein, and specific types of sugar to quantify the estimated effects of substituting the equivalent energy from sugars or starch for an equivalent percentage of energy from total fat. Additionally, we quantified the estimated effects of substituting energy from TGE or TFE for equivalent energy from vegetable and animal fat on CHD risk. We adjusted for additional covariates in multivariable models if they were a potential confounder, i.e., associated with diet and CHD risk. Our multivariable-adjusted models further included BMI (<21, 21-24.9, 25-29.9, 30-31.9, ≥32 kg/m²) [30, 31], race (non-Hispanic White or other races/ethnicities), physical activity (women: <3.0, 3.0-8.9, 9.0-17.9, 18.0-26.9, 27.0-41.9, ≥42.0; men: <3.0, 3.0-8.9, 9.0-17.9, 18.0-26.9, ≥27.0 MET hours/week), smoking status (never, former, current 1-14 cigarettes/d, current 15-24 cigarettes/d, current ≥25 cigarettes/d), menopausal status and hormone use in women (premenopausal, postmenopausal never users, postmenopausal past users, postmenopausal current users), alcohol consumption (women: 0, 0.1-4.9, 5.0-14.9, ≥15 g/d; men: 0, 0.1-4.9, 5.0-9.9, 10-14.9, 15-29.9, ≥30 g/d), family history of MI (yes or no), baseline history of hypertension or hypercholesterolemia, dietary cholesterol, and fiber intakes (from different food sources including cereal, fruit, vegetable, and legume). For repeatedly measured covariates, we included updated values as time-varying covariables in the model. Simple updating was used if a non-dietary covariate was not reported for a follow-up year. BMI measures were missing for <0.5% of participants. We conducted analyses in 2 cohorts separately and then pooled risk estimates using fixed-effects meta-analysis as there was no significant heterogeneity (P > 0.05) for the association of the primary exposures, TGE and TFE, with CHD for the 2 cohorts.

Associations of dietary carbohydrates with CHD in overweight individuals have been found to be more pronounced than in those with BMI <25 [6, 9]. Therefore, we conducted stratified analyses by BMI categories (<25, 25 to <30, ≥30 kg/m²) for TGE and TFE. Participants missing BMI measures (NHS, n = 90; HPFS, n = 869) were set to median values (NHS, BMI = 25 kg/m²; HPFS, BMI = 25.5 kg/m²) for stratified analyses. To address the possibility of an interaction between carbohydrate intake and alcohol intake, we conducted stratified analyses by alcohol intake categories (0 to <5, 5 to <15, ≥15 g/d) for TGE and TFE [19, 32]. We explored effect modification by age (<65 versus ≥65 y) and by physical activity level (<15 versus ≥15 MET-hr/wk), the approximate median in our population. All stratified analyses used time-varying data. To explore what variables may explain the difference in risk estimates between age-adjusted and multivariable-adjusted models, we conducted leave-one-out analyses. All data analyses were completed using SAS software (version 9.2, SAS Institute, Inc., Cary, North Carolina), and statistical significance was set at a 2-tailed P value < 0.05.

Results

Population characteristics

During up to 40 y of follow-up in 2 cohorts, we documented 9,723 incident CHD cases during 3,459,965 person-years of follow-up. In NHS (1980-2020), 4,641 CHD events, including 1,765 fatal cases, occurred during 2,407,676 person-years of follow-up. In HPFS (1986-2016), 5,082 CHD events, including 2,549 fatal cases, occurred during 1,052,289 person-years of follow-up.

At baseline, participants with higher total sugar, added sugar, TGE, and TFE intakes had lower intakes of alcohol and protein (Table 1). Those with higher intake of TGE and TFE were more likely to have high cholesterol and to consume more fruit fiber but less likely to be current smokers. Individuals with higher added sugar intake were less likely to be physically active and had lower protein intake.

TABLE 1.

Baseline characteristics according to quintiles of sugar intake

	Total Sugars			Added Sugar			Total Fructose Equivalents 1			Total Glucose Equivalents 2
	Q1	Q3	Q5	Q1	Q3	Q5	Q1	Q3	Q5	Q1	Q3	Q5
NHS (1980)
Sugar intake, % energy	12.4	22.8	37.2	3.3	9.1	21.0	5.1	9.8	17.3	18.4	26.2	34.9
Age, y	46.4	46.3	45.8	47.3	46.4	44.9	46.4	46.4	45.9	47.2	46.2	45.4
BMI, kg/m2	24.4	24.4	24.4	24.6	24.3	24.4	24.4	24.3	24.4	24.4	24.4	24.3
Alcohol, g/d	12.0	5.2	3.4	11.0	5.5	4.0	11.5	5.4	3.7	12.7	5.3	2.9
Physical activity, MET-h/wk	12.6	14.1	15.0	14.4	14.2	12.6	12.4	14.4	14.9	14.3	13.7	14.1
Current smoking, %	36.9	24.3	25.5	33.4	23.2	30.6	37.8	24.4	26.1	37.1	25.0	23.5
Premenopausal, %	55.4	56.4	54.4	55.1	56.8	54.8	55.6	56.4	54.5	54.7	56.0	54.9
Current menopausal hormone use, %	8.9	8.7	9.2	9.1	8.6	8.6	9.1	8.3	9.0	9.5	8.5	8.9
Family history of MI, %	9.1	9.6	8.9	9.1	9.4	9.2	8.8	9.5	9.1	8.8	9.6	9.6
Hypercholesterolemia, %	4.5	5.4	7.4	4.9	4.5	7.4	5.6	3.5	7.0	3.6	4.1	8.1
Hypertension, %	16.4	15.2	16.4	19.6	14.2	14.9	14.7	15.2	17.2	15.6	15.5	17.9
Total energy intake, kcal/d	1525	1581	1571	1469	1580	1639	1558	1575	1556	1528	1586	1556
Protein, % energy	20.5	19.4	17.2	21.5	19.4	16.3	20.8	19.4	16.8	20.6	19.2	17.4
Cereal fiber, g/d	2.2	2.7	2.3	2.2	2.7	2.3	2.3	2.7	2.3	2.0	2.6	2.7
Fruit fiber, g/d	2.3	4.4	6.0	4.1	4.5	3.9	2.1	4.3	6.3	3.2	4.3	5.3
Vegetable fiber, g/d	4.8	5.0	4.9	5.3	5.0	4.3	4.6	5.0	5.0	5.0	4.9	4.9
HPFS (1986)
Sugar intake, % energy	12.6	20.9	31.2	3.7	8.7	17.2	5.1	9.0	14.3	17.5	24.2	30.9
Age, y	52.6	53.1	53.2	54.0	53.1	51.7	52.8	53.2	53.1	54.6	52.7	51.9
BMI, kg/m2	25.9	25.4	25.2	25.7	25.5	25.3	25.9	25.4	25.2	26.0	25.5	24.8
Alcohol, g/d	22.3	9.7	5.0	18.2	10.6	6.8	21.2	9.8	5.5	24.5	9.0	4.1
Physical activity, MET-h/wk	18.4	21.1	22.9	22.0	21.3	19.1	18.1	20.9	22.9	18.8	20.7	23.1
Current smoking, %	14.0	8.9	7.6	9.8	8.9	11.8	14.9	8.3	7.8	14.9	8.8	6.4
Family history of MI, %	32.2	32.4	31.9	33.5	31.0	31.5	31.9	32.3	31.9	31.7	31.6	32.2
Hypercholesterolemia, %	10.6	10.0	10.8	11.4	9.7	9.3	9.9	10.2	11.3	10.0	9.6	12.0
Hypertension, %	21.6	18.9	19.0	22.6	18.7	18.1	21.1	19.2	19.4	22.7	18.8	18.0
Total energy intake, kcal/d	1906	2011	2044	1820	2004	2146	1930	2009	2027	1904	2018	2030
Protein, % energy	19.7	18.7	16.8	20.4	18.7	16.2	20.0	18.7	16.5	19.8	18.7	16.9
Cereal fiber, g/d	5.2	6.1	5.9	5.7	6.1	5.5	5.2	6.1	5.8	4.4	5.9	7.2
Fruit fiber, g/d	2.6	4.3	6.0	4.8	4.4	3.5	2.3	4.3	6.4	3.3	4.3	5.1
Vegetable fiber, g/d	6.6	7.0	6.9	7.7	7.1	5.8	6.3	7.0	7.1	6.3	6.9	7.5

Values are means for continuous variables and percentages for categorical variables. All variables except age were age standardized.

HPFS, Health Professionals Follow-up Study; MET, metabolic equivalent task; MI, myocardial infarction; NHS, Nurses’ Health Study; Q, quintile

¹Total fructose equivalents include fructose from sucrose.

²Total glucose equivalents include glucose from sucrose, lactose, maltose, and starch.

Pearson correlation between sugar types, fiber, starch, and glycemic load were assessed at the middle of follow-up in each cohort. TGE was highly correlated with starch intake (r = 0.75 in HPFS; r = 0.73 in NHS) and glycemic load (r = 0.84 in HPFS; r = 0.91 in NHS). TFE was highly correlated with total sugar (r = 0.93 in HPFS; r = 0.92 in NHS) and sucrose (r = 0.79 in HPFS; r = 0.82 in NHS) intake and moderately correlated with sugar intrinsic to food (0.53 in HPFS; r = 0.43 in NHS) (Supplemental Figure 2). TFE had a moderate positive correlation with fruit fiber (r = 0.46 in HPFS; r = 0.37 in NHS) and had almost no correlation with cereal, vegetable, or legume fiber (Supplemental Figure 3). Total carbohydrate intake was composed primarily of TGE (58%) followed by TFE (21%) and other forms (21%). Sources of dietary TFE were primarily from added sugar and juice (60%), followed by fruit and vegetable sources (40%).

In multivariable-adjusted models, higher intake of TGE, compared with total fat, was significantly associated with a higher risk of CHD. Comparing extreme quintiles of TGE intake, the HR (95% CI) of CHD was 1.31 (1.20 to 1.42; P _trend < 0.001) (Table 2). Replacing 5% of energy from total fat with equivalent energy from TGE was associated with a 12% higher CHD risk (HR: 1.12, 95% CI: 1.08 to 1.16) (Figure 1). As risk estimates differed between age-adjusted and multivariable-adjusted models, secondary analyses were conducted to identify variables that may explain the observed differences. Covariate adjustment for fiber intake was identified as a primary factor contributing to the observed association between TGE and higher CHD risk, as well as other associations, including total sugar and intrinsic sugar. Additionally, we explored the impact of the overall dietary pattern in secondary analyses by adjusting for the Alternative Healthy Eating Index, and the effect estimates for the primary exposures were similar. TFE intake was not significantly associated with CHD risk [quintile 5 (Q5) compared with Q1: 1.03 (95% CI: 0.94 to 1.11; P _trend = 0.25)] (Table 2). When we examined TFE intake from fruit and vegetables versus other sources, TFE intake from added sugar and fruit juice was significantly associated with CHD risk [Q5 compared with Q1: 1.12 (95% CI: 1.04 to 1.20; P _trend < 0.01)] whereas TFE intake from fruit and vegetable sources was not [Q5 compared with Q1: 0.98 (95% CI: 0.88 to 1.09; P _trend = 0.7)] (Supplemental Table 1).

TABLE 2.

Associations of total glucose equivalents and total fructose equivalents with coronary heart disease risk

	Quintiles of Sugar Intake					Ptrend	HR (95% CI) 1 5% increment energy
	Q1	Q2	Q3	Q4	Q5
Total glucose equivalents 2
NHS
Median (% energy)	22.8	27.1	29.3	31.3	34.1
Cases (n)	864	882	901	949	1045
Age-adjusted model 3	Ref.	0.98 (0.90, 1.08)	0.98 (0.89, 1.07)	0.99 (0.90, 1.08)	1.01 (0.92, 1.10)	0.87	1.00 (0.97, 1.04)
MV-adjusted model 4	Ref.	1.10 (0.99, 1.21)	1.16 (1.04, 1.29)	1.25 (1.12, 1.40)	1.39 (1.23, 1.57)	<0.001	1.16 (1.10, 1.22)
HPFS
Median (% energy)	20.8	25.1	27.7	29.9	33.2
Cases (n)	971	1002	1020	1048	1041
Age-adjusted model 3	Ref.	1.01 (0.92, 1.10)	1.01 (0.92, 1.10)	1.01 (0.92, 1.10)	0.99 (0.91, 1.08)	0.82	1.00 (0.96, 1.03)
MV-adjusted model 4	Ref.	1.10 (1.00, 1.21)	1.15 (1.04, 1.28)	1.19 (1.06, 1.33)	1.23 (1.09, 1.39)	<0.001	1.09 (1.04, 1.14)
Pooled 5
Age-adjusted model 3	Ref.	1.00 (0.93, 1.06)	0.99 (0.93, 1.06)	1.00 (0.94, 1.06)	1.00 (0.94, 1.06)	0.95	1.00 (0.97, 1.02)
MV-adjusted model 4	Ref.	1.10 (1.02, 1.18)	1.15 (1.07, 1.24)	1.22 (1.13, 1.32)	1.31 (1.20, 1.42)	<0.001	1.12 (1.08, 1.16)
Total fructose equivalents 6
NHS
Median (% energy)	6.5	8.4	9.7	11.2	13.6
Cases (n)	890	893	842	959	1057
Age-adjusted model 3	Ref.	0.92 (0.84, 1.01)	0.80 (0.73, 0.88)	0.84 (0.77, 0.92)	0.87 (0.79, 0.95)	0.002	0.91 (0.86, 0.97)
MV-adjusted model 4	Ref.	0.99 (0.90, 1.10)	0.90 (0.81, 0.99)	0.97 (0.87, 1.08)	0.99 (0.88, 1.11)	0.96	1.00 (0.92, 1.08)
HPFS
Median (% energy)	6.0	7.9	9.3	10.8	13.3
Cases (n)	952	914	979	1047	1190
Age-adjusted model 3	Ref.	0.87 (0.79, 0.95)	0.87 (0.79, 0.95)	0.88 (0.81, 0.96)	0.99 (0.91, 1.08)	0.63	1.01 (0.96, 1.07)
MV-adjusted model 4	Ref.	0.92 (0.83, 1.01)	0.94 (0.85, 1.04)	0.97 (0.87, 1.08)	1.06 (0.94, 1.19)	0.10	1.07 (0.99, 1.15)
Pooled 5
Age-adjusted model 3	Ref.	0.89 (0.83, 0.95)	0.83 (0.78, 0.89)	0.86 (0.81, 0.92)	0.93 (0.87, 0.99)	0.08	0.96 (0.93, 1.00)
MV-adjusted model 4	Ref.	0.96 (0.89, 1.02)	0.92 (0.85, 0.99)	0.97 (0.90, 1.05)	1.03 (0.94, 1.11)	0.25	1.03 (0.98, 1.09)

Data are represented as HR (95% CIs) estimated with Cox proportional hazards models unless otherwise indicated. Models represent the effect of substituting a percentage of energy from individual sugars for an equivalent percentage of energy from total fat.

CI, confidence interval; HPFS, Health Professionals Follow-up Study; HR, hazard ratio; MV, multivariable; NHS, Nurses’ Health Study; Q, quintile

¹HR (95% CI) of coronary heart disease for substituting 5% of energy from total fructose equivalents and total glucose equivalents for the same energy from total fat.

²Total glucose equivalents include glucose from sucrose, lactose, maltose, and starch.

³Age-adjusted model adjusted for age (in months)

⁴Multivariable-adjusted model adjusted for: Age (in month), non-Hispanic White (yes vs. no), physical activity (women: <3.0, 3.0-8.9, 9.0-17.9, 18.0-26.9, 27.0-41.9, ≥42.0; men: <3.0, 3.0-8.9, 9.0-17.9, 18.0-26.9, ≥27.0 h of metabolic equivalent tasks per week), smoking status (never, past, current 1-14 cigarettes/d, current 15-24 cigarettes/d, current ≥25 cigarettes/d), alcohol consumption (women: 0, 0.1-4.9, 5.0-14.9, ≥15 g/d; men: 0, 0.1-4.9, 5.0-9.9, 10-14.9, 15-29.9, ≥30 g/d), family history of myocardial infarction (yes vs. no), menopausal status and hormone use in women (premenopausal, postmenopausal never users, postmenopausal past users, postmenopausal current users), BMI (<21, 21-24.9, 25-29.9, 30-31.9, ≥32 kg/m²), baseline histories of hypertension and hypercholesterolemia (yes vs. no), intakes of total energy, dietary cholesterol, percentage of energy intake from dietary protein, and intakes of cereal fiber, fruit fiber, vegetable fiber, and legume fiber. Total fructose equivalents and total glucose equivalents models also include percentages of energy intake from remaining carbohydrate intake (all in quintiles).

⁵Pooled results were calculated using the fixed-effects model from NHS and HPFS.

⁶Total fructose equivalents include fructose from sucrose.

FIGURE 1.

Estimated effects of isocaloric substitution of total glucose equivalents or total fructose equivalents for total fat on the risk of coronary heart disease.

Change in the risk of coronary heart disease with isocaloric substitution of 5% energy from total glucose equivalents, total fructose equivalents, TFE from fruit and vegetables, and TFE from added sugar and juice for equivalent energy from dietary total fat. Change in risk is derived from hazard ratios and represented as solid bars; error bars represent 95% confidence intervals (∗P < 0.05). The multivariate models were adjusted for age, non-Hispanic White (yes vs. no), physical activity, smoking status, alcohol consumption, menopausal status and hormone use in women, BMI, baseline histories of hypertension and hypercholesterolemia, percent energy intake from remaining carbohydrate intake. Hazard ratios were calculated with Cox proportional hazards models and pooled using fixed-effects models from NHS and HPFS. HPFS, Health Professionals Follow-up Study; TFE, total fructose equivalents; F&V, fruits and vegetables; NHS, Nurses’ Health Study

Total sugar, added sugar, and starch were also significantly associated with a higher risk of CHD. Comparing extreme quintiles, HRs of CHD risk were 1.16 (95% CI: 1.07 to 1.26; P _trend < 0.001) for total sugar intake and 1.08 (95% CI: 0.99 to 1.16; P _trend = 0.04) for added sugar intake (Table 3). Replacing 5% of energy from total fat with equivalent energy from total and added sugars was associated with a higher CHD risk of 5% (HR: 1.05, 95% CI: 1.02 to 1.08) and 4% (HR 1.04, 95% CI: 1.00 to 1.07). Comparing extreme quintiles of starch intake, HRs of CHD risk were 1.24 (95% CI: 1.14 to 1.34; P _trend < 0.001) (Table 3). Replacing 5% energy from total fat with energy from starch was associated with an 11% higher CHD risk (HR: 1.11, 95% CI: 1.07 to 1.15). Intakes of sucrose and lactose were not associated with CHD (Table 3). Replacing 5% energy from total fat with energy from sugar intrinsic to foods was associated with a 5% higher CHD risk (HR: 1.05, 95% CI: 1.01 to 1.09) (Table 3). In stratified analyses, associations of TGE and TFE with CHD risk were generally consistent across categories of BMI status, age, alcohol intake, and physical activity (Supplemental Table 2).

TABLE 3.

Associations of different types of sugar/starch with coronary heart disease risk

	Quintiles of Sugar/Starch Intake					Ptrend	HR (95% CI) 1 5% increment energy
	Q1	Q2	Q3	Q4	Q5
Total sugar
NHS
Median (% energy)	15.9	20.3	23.3	26.4	31.2
Cases (n)	849	865	892	975	1060
Age-adjusted model 2	Ref.	0.93 (0.85, 1.03)	0.89 (0.81, 0.98)	0.90 (0.82, 0.99)	0.90 (0.82, 0.98)	0.02	0.97 (0.94, 1.00)
MV-adjusted model 3	Ref.	1.04 (0.94, 1.15)	1.05 (0.95, 1.17)	1.12 (1.01, 1.24)	1.17 (1.04, 1.31)	0.004	1.05 (1.02, 1.09)
HPFS
Median (% energy)	14.5	18.7	21.7	24.8	29.6
Cases (n)	885	944	982	1066	1205
Age-adjusted model 2	Ref.	0.97 (0.88, 1.06)	0.93 (0.85, 1.02)	0.95 (0.86, 1.03)	1.04 (0.95, 1.13)	0.35	1.01 (0.99, 1.04)
MV-adjusted model 3	Ref.	1.03 (0.94, 1.13)	1.02 (0.92, 1.13)	1.05 (0.95, 1.17)	1.16 (1.03, 1.30)	0.01	1.05 (1.01, 1.09)
Pooled 4
Age-adjusted model 2	Ref.	0.95 (0.89, 1.02)	0.91 (0.85, 0.97)	0.92 (0.86, 0.98)	0.97 (0.91, 1.03)	0.35	0.99 (0.97, 1.01)
MV-adjusted model 3	Ref.	1.03 (0.97, 1.11)	1.04 (0.96, 1.11)	1.08 (1.01, 1.17)	1.16 (1.07, 1.26)	<0.001	1.05 (1.02, 1.08)
Added sugar
NHS
Median (% energy)	5.2	7.6	9.5	11.8	16.2
Cases (n)	993	889	875	860	1024
Age-adjusted model 2	Ref.	0.89 (0.81, 0.98)	0.86 (0.79, 0.94)	0.82 (0.75, 0.90)	0.99 (0.91, 1.08)	1.00	1.00 (0.96, 1.04)
MV-adjusted model 3	Ref.	0.93 (0.85, 1.02)	0.92 (0.83, 1.01)	0.87 (0.79, 0.97)	1.03 (0.92, 1.15)	0.60	1.01 (0.97, 1.06)
HPFS
Median (% energy)	4.7	7.1	9.0	11.3	15.5
Cases (n)	966	948	1018	1026	1124
Age-adjusted model 2	Ref.	0.94 (0.86, 1.03)	1.01 (0.92, 1.10)	1.00 (0.91, 1.09)	1.17 (1.08, 1.28)	<0.001	1.09 (1.05, 1.13)
MV-adjusted model 3	Ref.	0.97 (0.88, 1.06)	1.03 (0.94, 1.13)	1.00 (0.91, 1.11)	1.13 (1.01, 1.26)	0.02	1.06 (1.01, 1.11)
Pooled 4
Age-adjusted model 2	Ref.	0.92 (0.86, 0.98)	0.93 (0.88, 0.99)	0.91 (0.85, 0.97)	1.08 (1.02, 1.15)	0.002	1.04 (1.02, 1.07)
MV-adjusted model 3	Ref.	0.95 (0.89, 1.01)	0.97 (0.91, 1.04)	0.94 (0.87, 1.01)	1.08 (0.99, 1.16)	0.04	1.04 (1.00, 1.07)
Intrinsic sugar 5
NHS
Median (% energy)	7.9	11.0	13.3	15.7	19.4
Cases (n)	864	832	884	986	1075
Age-adjusted model 2	Ref.	0.86 (0.78, 0.95)	0.84 (0.76, 0.92)	0.85 (0.77, 0.93)	0.81 (0.74, 0.88)	<0.001	0.92 (0.89, 0.96)
MV-adjusted model 3	Ref.	1.04 (0.94, 1.15)	1.11 (0.99, 1.24)	1.20 (1.07, 1.34)	1.22 (1.08, 1.38)	<0.001	1.10 (1.04, 1.16)
HPFS
Median (% energy)	7.0	9.8	12.0	14.4	18.4
Cases (n)	963	901	986	1036	1196
Age-adjusted model 2	Ref.	0.80 (0.73, 0.88)	0.80 (0.73, 0.88)	0.78 (0.71, 0.85)	0.79 (0.72, 0.86)	<0.001	0.92 (0.89, 0.96)
MV-adjusted model 3	Ref.	0.87 (0.79, 0.96)	0.90 (0.81, 1.00)	0.90 (0.80, 1.01)	0.96 (0.84, 1.08)	1.00	1.00 (0.95, 1.06)
Pooled 4
Age-adjusted model 2	Ref.	0.83 (0.78, 0.89)	0.82 (0.77, 0.87)	0.81 (0.76, 0.86)	0.80 (0.75, 0.85)	<0.001	0.92 (0.90, 0.95)
MV-adjusted model 3	Ref.	0.95 (0.88, 1.02)	1.00 (0.92, 1.07)	1.04 (0.96, 1.12)	1.08 (0.99, 1.18)	0.01	1.05 (1.01, 1.09)
Sucrose
NHS
Median (% energy)	5.7	7.7	9.1	10.7	13.3
Cases (n)	974	883	905	855	1024
Age-adjusted model 2	Ref.	0.86 (0.79, 0.94)	0.83 (0.76, 0.91)	0.74 (0.68, 0.81)	0.84 (0.77, 0.91)	<0.001	0.88 (0.84, 0.93)
MV-adjusted model 3	Ref.	0.91 (0.83, 1.00)	0.89 (0.81, 0.98)	0.80 (0.72, 0.88)	0.87 (0.78, 0.97)	<0.01	0.91 (0.85, 0.97)
HPFS
Median (% energy)	5.2	7.1	8.4	9.9	12.4
Cases (n)	894	936	976	1054	1222
Age-adjusted model 2	Ref.	0.96 (0.88, 1.06)	0.95 (0.86, 1.04)	0.96 (0.87, 1.05)	1.06 (0.97, 1.16)	0.12	1.05 (0.99, 1.11)
MV-adjusted model 3	Ref.	1.01 (0.92, 1.11)	0.99 (0.90, 1.09)	0.99 (0.90, 1.10)	1.05 (0.95, 1.17)	0.34	1.03 (0.96, 1.11)
Pooled 4
Age-adjusted model 2	Ref.	0.91 (0.85, 0.97)	0.89 (0.83, 0.94)	0.85 (0.79, 0.90)	0.94 (0.89, 1.00)	0.05	0.96 (0.92, 1.00)
MV-adjusted model 3	Ref.	0.96 (0.90, 1.02)	0.94 (0.88, 1.01)	0.89 (0.83, 0.96)	0.96 (0.89, 1.03)	0.17	0.97 (0.92, 1.01)
Lactose
NHS
Median (% energy)	1.0	2.1	3.2	4.4	7.1
Cases (n)	987	836	862	923	1033
Age-adjusted model 2	Ref.	0.85 (0.77, 0.93)	0.85 (0.78, 0.93)	0.89 (0.81, 0.97)	0.94 (0.86, 1.02)	0.80	0.99 (0.93, 1.06)
MV-adjusted model 3	Ref.	0.90 (0.82, 0.99)	0.92 (0.84, 1.01)	0.96 (0.87, 1.05)	1.00 (0.91, 1.11)	0.37	1.04 (0.96, 1.12)
HPFS
Median (% energy)	0.7	1.5	2.4	3.4	5.9
Cases (n)	1017	1021	952	992	1100
Age-adjusted model 2	Ref.	0.97 (0.89, 1.06)	0.86 (0.79, 0.94)	0.88 (0.80, 0.96)	0.94 (0.86, 1.02)	0.14	0.94 (0.87, 1.02)
MV-adjusted model 3	Ref.	0.99 (0.91, 1.08)	0.89 (0.81, 0.98)	0.91 (0.83, 1.00)	0.94 (0.85, 1.03)	0.15	0.94 (0.87, 1.02)
Pooled 4
Age-adjusted model 2	Ref.	0.91 (0.86, 0.97)	0.86 (0.80, 0.91)	0.88 (0.83, 0.94)	0.94 (0.88, 1.00)	0.24	0.97 (0.92, 1.02)
MV-adjusted model 3	Ref.	0.95 (0.89, 1.01)	0.91 (0.85, 0.97)	0.93 (0.87, 1.00)	0.97 (0.91, 1.04)	0.74	0.99 (0.94, 1.05)
Starch
NHS
Median (% energy)	11.8	15.0	16.9	18.6	21.2
Cases (n)	967	917	868	923	966
Age-adjusted model 2	Ref.	0.97 (0.89, 1.07)	0.94 (0.86, 1.03)	1.01 (0.92, 1.10)	1.04 (0.96, 1.14)	0.29	1.02 (0.98, 1.07)
MV-adjusted model 3	Ref.	1.10 (1.00, 1.21)	1.12 (1.01, 1.23)	1.23 (1.11, 1.37)	1.33 (1.18, 1.49)	<0.001	1.16 (1.10, 1.23)
HPFS
Median (% energy)	11.1	14.4	16.4	18.3	21.4
Cases (n)	1117	1034	1041	923	967
Age-adjusted model 2	Ref.	0.94 (0.86, 1.02)	0.98 (0.90, 1.06)	0.89 (0.81, 0.97)	0.95 (0.87, 1.03)	0.10	0.97 (0.93, 1.01)
MV-adjusted model 3	Ref.	1.02 (0.93, 1.12)	1.11 (1.01, 1.22)	1.04 (0.94, 1.15)	1.16 (1.04, 1.30)	0.01	1.07 (1.01, 1.12)
Pooled 4
Age-adjusted model 2	Ref.	0.95 (0.90, 1.02)	0.96 (0.90, 1.02)	0.94 (0.89, 1.00)	0.99 (0.93, 1.06)	0.59	0.99 (0.96, 1.02)
MV-adjusted model 3	Ref.	1.06 (0.99, 1.13)	1.11 (1.04, 1.19)	1.13 (1.05, 1.21)	1.24 (1.14, 1.34)	<0.001	1.11 (1.07, 1.15)

Data are represented as HR (95% CIs) unless otherwise indicated and are estimated with Cox proportional hazards models. Models represent the effect of substituting a percentage of energy from individual sugars for an equivalent percentage of energy from total fat.

CI, confidence interval; HPFS, Health Professionals Follow-up Study; HR, hazard ratio; MV, multivariable; NHS, Nurses’ Health Study; Q, quintile

¹HR (95% CI) of coronary heart disease for substituting 5% of energy from total sugar, added sugar, intrinsic sugar, sucrose, lactose, and starch for the same energy from total fat.

²Age-adjusted model adjusted for age (in months)

³Multivariable-adjusted model adjusted for: Age (in months), non-Hispanic White (yes vs. no), physical activity (women: <3.0, 3.0-8.9, 9.0-17.9, 18.0-26.9, 27.0-41.9, ≥42.0; men: <3.0, 3.0-8.9, 9.0-17.9, 18.0-26.9, ≥27.0 h of metabolic equivalent tasks per week), smoking status (never, past, current 1-14 cigarettes/d, current 15-24 cigarettes/d, current ≥25 cigarettes/d), alcohol consumption (women: 0, 0.1-4.9, 5.0-14.9, ≥15 g/d; men: 0, 0.1-4.9, 5.0-9.9, 10-14.9, 15-29.9, ≥30 g/d), family history of myocardial infarction (yes vs. no), menopausal status and hormone use in women (premenopausal, postmenopausal never users, postmenopausal past users, postmenopausal current users), BMI (<21, 21-24.9, 25-29.9, 30-31.9, ≥32 kg/m²), baseline histories of hypertension and hypercholesterolemia (yes vs. no), intakes of total energy, dietary cholesterol, percentage of energy intake from dietary protein, and intakes of cereal fiber, fruit fiber, vegetable fiber, and legume fiber. Total sugar, added sugar, intrinsic sugar, sucrose, lactose, and starch models also include percentages of energy intake from remaining carbohydrate intake (all in quintiles).

⁴Pooled results were calculated using fixed-effects models from NHS and HPFS.

⁵Sugar intrinsic to foods (e.g., sugars from fruit and 100% fruit juices, but not added sugar).

In additional substitution analyses, substituting 5% of energy from TGE with equivalent energy from vegetable fat was significantly associated with a 16% decrease in CHD risk (HR: 0.84, 95% CI: 0.80 to 0.87) (Supplemental Figure 4), whereas substituting animal fat was associated with a 6% decrease in CHD risk (HR: 0.94, 95% CI: 0.90 to 0.98). Similarly, substituting 5% of energy from TFE with equivalent energy from vegetable fat was associated with a 9% decrease in CHD risk (HR: 0.91, 95% CI: 0.86 to 0.97), and substituting animal fat was not associated with decreased risk (Supplemental Figure 4).

Discussion

In two large cohorts of US men and women with extensive follow-up, we found higher intakes of TGE, TFE from added sugar and juice, total sugar, added sugar, and starch were associated with a higher risk of CHD, while TFE from fruits and vegetables, TFE, sucrose, and lactose were not significantly associated with CHD risk. In our estimates, the isocaloric substitution of total fat with TGE and TFE from added sugar and juice was associated with higher CHD risk. These findings are consistent with previous cross-sectional analyses in these cohorts, linking SSB and juice, but not fruit, with circulating levels of several cardiometabolic biomarkers [33].

Previous population studies indicate dietary carbohydrate quality is a more important determinant of CHD risk than total carbohydrate intake [1, 9, 34]. Compared with these earlier studies on glycemic index/glycemic load that evaluated carbohydrate quality based on a single biological pathway, i.e., the postprandial glycemic response, our study captured the heterogeneous metabolic effects of different sugar types through multiple potential biological pathways. Similar to our findings, measures of carbohydrate quality, including glycemic load and individual carbohydrate types, such as starch and added sugar, have been associated with higher CHD risk [1, 9, [34], [35], [36], [37]]. Recommended intakes for added sugars (Dietary Guidelines for Americans, 2020-2025) and free sugars (World Health Organization) are less than 10% of total calories [19, 38]. In our population, individuals in the highest quintile of intake consumed about 16% of energy from added sugar, which was associated with a slightly higher CHD risk; when TGE was considered, a variable including starch and added sugar contributions to glucose intake, CHD risk was 30% higher. In a meta-analysis of prospective cohort studies, glycemic index and glycemic load were associated with 26% and 55% higher CHD events, respectively, among women [34]. In concordance with our findings, two prospective cohort studies found harmful associations of added sugar intake with cardiovascular disease (CVD) mortality and CHD risk [1, 6]; in the NHANES study, participants in the highest intake quintile for added sugar (21% of energy intake) had a two-fold higher risk of CVD mortality [6]. However, results from studies of starch intake have been mixed, with some showing no association [39], whereas others show significant associations [1] or associations only among a specific subgroup [2]. A previous study among Chinese individuals found higher starch intake was associated with about a 60% greater risk of CHD among women within the highest quintile of starch intake, which was about 54% of total energy [2]. Our study indicates a significant association between starch intake and CHD risk and additionally suggests that the combined influence of all glucose contributors via TGE in the diet (e.g., starch and sucrose) may be important to consider when assessing CHD risk.

Prospective studies investigating the specific role of fructose on CHD risk factors or incident CHD have been limited. A recent meta-analysis of food sources containing fructose sugars and incident hypertension showed harmful associations with SSB intake but protective associations with fruit, yogurt, and moderate doses of 100% fruit juice [40]. These findings align with other research on fructose food sources and CHD risk, with consistent harmful associations for regular SSB intake and protective associations with fruit intake [3, 4, [41], [42], [43]]. Similar findings suggest that drinking as little as one SSB per day can lower HDL cholesterol and raise triglycerides [44]. Fructose metabolism within the liver may vary by food source, particularly when consumed as excess calories from added sugars [45, 46]; this can result in excess de novo lipogenesis, increasing circulating triglycerides.

Most previous population studies capture fructose exposures as a component of sucrose, added sugar, or fruit intake. A recent dose-response meta-analysis of sugars found no association between fructose intake and CVD incidence and a slight but nonsignificant positive association between sucrose intake and CVD incidence [7]. In a systematic review of randomized controlled trials of fructose, the impact of fructose on blood lipids was significant only in the context of excess calories (+21 to 35% energy) [47]. Although some clinical studies have demonstrated links between fructose and CHD risk factors like triglycerides and LDL cholesterol, we did not observe a significant association between overall TFE intake and CHD risk [47]. However, TFE from sources other than fruits and vegetables, particularly SSBs and other added sugars, was associated with higher CHD risk. In contrast, we found no evidence for an association of TFE from fruits and vegetables. Fruit is a major source of fructose, yet fruit is consistently associated with beneficial effects on CHD risk [[41], [42], [43], 48]. This may be due to the cardioprotective impact of polyphenols, potassium, and other nutrients within fruits [48], which have antioxidant and anti-inflammatory properties that can protect blood vessels and overall vascular function. Additionally, the fiber content and food matrix of whole fruits and vegetables result in a much slower rate of digestion and absorption of fructose than that from fruit juices and other beverages [49]. About 40% of total TFE was from fruit and vegetable sources. Similarly, the lack of a significant association of total sucrose with CHD may be due to the contribution of fruit to sucrose intake (e.g., most of the sugar in mango, cantaloupe, and oranges is in the form of sucrose).

The strong association we found between TGE intake and CHD risk may be due to a variety of potential mechanisms. TGE captures the combined influence of glucose from dietary starch and added sugars, making up almost 35% of energy intake among individuals within the highest quintile of consumption. Although whole grains contribute to starch intake, whole grain consumption is protective for CVD outcomes, and intake continues to be below recommended levels [50]. Adjustment for fiber intake was therefore important to reveal the association of TGE with CHD risk more clearly. Data from the 2015-2016 NHANES survey estimated energy intake from starch sources, including refined grains, potatoes, and other starchy vegetables, at about 18% of total energy intake in US adults [51], making starch an important contributor to overall glucose load. Modern starchy foods, which are primarily refined starches, lead to high glycemic responses, and the accumulated impact of these physiological responses may increase risk of CHD.

Total sucrose was not associated with CHD risk. Similar to fructose, clinical studies of sucrose as part of isocaloric diets have found either no or limited effects on cardiometabolic risk factors [[52], [53], [54]]. In a study of eucaloric high versus low sucrose diets, no detrimental effects on glycemic profiles or vascular compliance were found [52].

Study strengths and limitations

Our study had both strengths and limitations. As with all observational studies, we cannot prove causality or rule out the potential influence of residual confounding. Our measures of individual sugars cannot account for the influence of the overall food matrix. However, the results remained largely unchanged after further adjustment for overall dietary pattern. Another limitation is the individual food sources, and glycemic index/load are likely to be important but are not captured here. Study participants were primarily Caucasian health professionals, which may limit the generalizability of our findings to other populations. Our study also has multiple important strengths, including large sample size, decades of follow-up with a large number of CHD events, and repeated assessments of diet and lifestyle using extensively validated tools. Additionally, our study yielded results consistent across two cohorts, with multiple repeated dietary assessments and careful adjustment for many potential time-varying confounding variables. We used cumulative average measures of diet to best represent long-term dietary patterns and minimize measurement error. Our models included extensive adjustment for confounding factors to improve the generalizability of the effect estimates to other populations and more closely reflect an estimate of the underlying physiological mechanisms relating sugar intake to CHD risk.

In conclusion, our findings support recommendations to reduce intake of TGE and TFE from added sugar and juice for CHD prevention. Total TFE intake and TFE from fruits and vegetables were not associated with CHD risk. Of total TFE intake, over half was from sources of added sugar and juice. TGE primarily reflects the intake of added sugar and starch. Reducing TFE from nonfruit and vegetable sources and reducing TGE from starch and added sugar is important for CHD prevention.

Acknowledgments

We would like to thank the participants and staff from NHS and HPFS for their valuable contributions.

Author contribution

The authors’ responsibilities were as follows—KKD, MJS, and DDW contributed to the study design and statistical methods. YL provided statistical expertise. KKD conducted analyses and drafted the manuscript. All authors contributed to the interpretation of results and critical revision of the manuscript for intellectual content. All authors read and approved the final manuscript.

Conflict of Interest

The authors report no conflicts of interest.

Funding

This study is supported by the National Institutes of Health grants UM1 CA186107, U01 CA167552, R01 HL034594, R01 HL035464, R01 NR019992, R00 DK119412, and T32 CA009001. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders had no role in considering the study design or in the collection, analysis, interpretation of data, writing of the report, or decision to submit the article for publication.

Data availability

Further information, including the procedures to obtain and access data from the Nurses’ Health Studies and Health Professionals Follow-up Study, is described at https://nurseshealthstudy.org/researchers. Because of participant confidentiality and privacy concerns, data are available upon reasonable written request.

Appendix A. Supplementary data

The following is the Supplementary data to this article: