Relationship Between Carbohydrate Intake (Quantity, Quality, and Time Eaten) and Mortality (Total, Cardiovascular, and Diabetes): Assessment of 2003–2014 National Health and Nutrition Examination Survey Participants

Poor diet is considered the major cause of death and the third main cause of disability-adjusted life-year loss in the U.S. (1). Current evidence indicates that in addition to the quantity of dietary intake, the quality of the overall diet also plays a critical role in maintaining health (2,3). More importantly, recent studies have shown that the timing of food intake is as important as the quantity and quality of the food (4–6). However, few studies have included examination of whether food quantity, quality, or intake timing is more important for health.

Generally, dietary carbohydrates provide almost 50% of total energy and are the main source of energy throughout the day (2). In the past 20 years, total carbohydrate intake during a day has remained stable among U.S. adults, whereas the quality of carbohydrates has slowly increased (7), probably because of dietary guidelines and media publicity based on the evidence of the beneficial effect of high-quality carbohydrates intake. Furthermore, investigators of a few recent studies found that the timing of macronutrient intake, independent of total daily carbohydrate intake, is associated with incidence of obesity (6,8–10), dyslipidemia (9,11), hyperglycemia (12), metabolic syndrome (6,10), and mortality (13). However, few researchers have examined the health effects of carbohydrates by considering the overall quantity, quality, and timing of intake simultaneously. Based on the above evidence, we hypothesize that the timing of carbohydrate intake is as important as the quality and quantity of carbohydrates. This hypothesis means that people who consume a large amount of high-quality carbohydrates each day may further improve their long-term survival by ensuring the correct timing of carbohydrate consumption, and people who consume a large amount of low-quality carbohydrates each day may further reduce their long-term survival by having an undesirable timing of carbohydrate consumption irrespective of the total daily consumption of carbohydrates.

To examine this hypothesis, we explored the association of timing of carbohydrate intake with all-cause mortality, cardiovascular disease (CVD) mortality, and diabetes mortality among people with low- and high-quality carbohydrate intake throughout the day using the U.S. National Health and Nutrition Examination Survey (NHANES) (2003–2014).

This study involved the data of adult participants of NHANES (2003–2014) aged >18 years who completed at least one valid dietary recall. Participants with extreme energy consumption (<500 or >3,500 kcal/day for women and <800 or >4,200 kcal/day for men), participants without information on the number of person-months of follow-up from the NHANES interview date, and pregnant women were excluded. Overall, 27,623 participants were included. NHANES is a stratified and multistage study conducted with a nationally representative sample of the U.S. population (14), the detailed descriptions of which are provided elsewhere (7). The NHANES protocol was approved by the National Center for Health Statistics Research Ethics Review Board, and all participants provided informed consent.

Participants’ food intake on two nonconsecutive days was collected during two 24-h dietary recall interviews. During the interview, the participants were asked to report the intake time of each food and beverage item and also asked to choose the name of the eating event: breakfast, lunch, dinner, supper, or snack. The first interview was conducted in person, and the second was conducted by telephone 3–10 days later. Dietary information was derived from the U.S. Department of Agriculture Food and Nutrient Database for Dietary Studies (FNDDS). The average food intake during the 2 days was calculated. Based on the MyPyramid Equivalents Database, 2.0 for USDA Survey Foods, 2003–2004 (MPED 2.0), the dietary intake of NHANES participants was integrated into 37 MyPyramid major groups. The timing for eating breakfast was set as at or after 4:00 a.m. and before 11:00 a.m., lunch at or after 11:00 a.m. and before 4:00 p.m., and dinner at or after 4:00 p.m. and before 10:00 p.m.

The main exposures were the quantity, quality, and daily timing of carbohydrate intake. The total daily intake of high- and low-quality carbohydrates was calculated with the method described in a previous study (7). The total daily high-quality carbohydrates intake was calculated based on the sum of the intake amount of whole grains, legumes, whole fruits, and nonstarchy vegetables, and the total daily low-quality carbohydrates intake was calculated based on the sum of the intake amount of refined grains, fruit juices, starchy vegetables, and added sugars. Detailed serving sizes are shown in Supplementary Table 1. Total daily intake amounts of high- and low-quality carbohydrates were two independent variables. The daily timing of carbohydrate intake was assessed with the calculation of the differences between the intake at dinner and breakfast (Δ = dinner − breakfast). The differences (Δ) for high- and low-quality carbohydrates intake between dinner and breakfast were calculated.

The outcomes in the current study were mortality that occurred after the survey participation date and before 31 December 2015, based on National Death Index data. ICD-10 was adopted to define cause-specific mortality. Mortality from CVD was determined for participants with ICD-10 codes of I00-I09, I11, I13, I20–I51, or I60–I69. Mortality from diabetes was determined for participants with codes of E10–E14. Overall, 2,545 deaths were documented, including 760 CVD deaths and 279 diabetes deaths.

Potential covariates included age (years), race (non-Hispanic White/non-Hispanic Black/Mexican American/other), sex (male/female), annual household income (<$20,000/$20,000–$45,000/$45,000–$75,000/>$100,000), education level (<9th grade/9th–11th grade/high school graduate/GED or equivalent/college or Associate of Arts degree/college graduate or higher), regular exercise (yes/no), BMI, smoking (yes/no), alcohol consumption (yes/no), self-reported hypertension, self-reported diabetes, self-reported hypercholesteremia, chronic diseases (ever diagnosed with congestive heart failure/coronary heart disease/angina/heart attack/stroke/cancer), dietary supplement use (yes/no), total energy intake (kilocalories per day), total daily high-quality carbohydrates intake (servings per day), total daily low-quality carbohydrates intake (servings per day), saturated fatty acids (SFAs) intake (grams per day), skipping breakfast, overall dietary quality calculated with the Alternative Healthy Eating Index (AHEI) score (15), and timing of breakfast and dinner.

Absolute food consumption was adjusted for total energy intake with the residual method to correct for measurement errors (16). Data on demographic characteristics, anthropometric measurements, and dietary intake as continuous variables are shown as the means (SEs) and for categorical variables are shown as numbers (percentages). General linear models were adopted to compare continuous baseline characteristics, with adjustment for age, and χ² tests were used to test for categorical variables.

The total daily carbohydrate intake as well as the daily high- and low-quality carbohydrate intake were categorized into tertiles. Cox proportional hazards (CPH) regression models were used to evaluate the association between the three variables and total, CVD, and diabetes mortality. Survival time is described in months between the NHANES interview date and death or census date (31 December 2015). Moreover, for determination of whether quantity, quality, or timing of carbohydrate intake is more important for long-term survival, for this study we adopted the analysis strategy by categorizing the participants according to daily carbohydrate intake patterns, which was established with the two variables of daily high- and low-quality carbohydrate intake. With each variable, the participants were categorized into two groups, high median value versus low median value, which generated four patterns of daily carbohydrate intake, as shown in Fig. 1. The Δhigh- and Δlow-quality carbohydrate intake between dinner and breakfast were categorized into tertiles. The association of Δhigh- and Δlow-quality carbohydrate intake with total, CVD, and diabetes mortality among the four carbohydrate intake patterns was evaluated with CPH regression models. A series of covariates, including age, race, sex, income, education, regular exercise, BMI, smoking, alcohol consumption, self-reported hypertension, self-reported diabetes, self-reported hypercholesteremia, chronic diseases (ever diagnosed with CVD/cancer), dietary supplement use, total energy intake, total daily high-quality carbohydrates intake (when the exposure was total daily low-quality carbohydrates intake or Δhigh- or Δlow-quality carbohydrates intake), total daily low-quality carbohydrates intake (when the exposure was total daily high-quality carbohydrates intake or Δhigh- and Δlow-quality carbohydrates intake), SFAs intake, skipping breakfast, and AHEI score, were also adjusted for in these CPH regression models.

In this study we also developed a series of predicted isocaloric models to assess the extent to which replacing 1 serving low-quality carbohydrates at dinner with 1 serving high-quality carbohydrates would impact total, CVD, and diabetes mortality by holding total daily energy as well as total daily high- and low-quality carbohydrate intake constant. Low-quality and high-quality carbohydrate intake at dinner was included in the models as continuous variables simultaneously. For each replacement of low-quality with high-quality carbohydrates intake, the difference between the β-coefficients of the two variables was adopted to estimate the hazard ratio (HR) and the variances and covariance of the two variables were adopted to estimate the 95% CI (17).

A two-sided P value <0.05 was considered statistically significant. All analyses were performed with R 3.6.1 (www.r-project.org/).

Seven sensitivity analyses were conducted to evaluate the robustness of the results: 1) analyzing the differences for food groups consumed at dinner and breakfast (Δfoods = dinner − breakfast); 2) excluding participants who skipped breakfast, which is a traditional dietary risk factor (18); 3) excluding participants who skipped dinner; 4) excluding participants with a follow-up time of <2 years or who died within 2 years of follow-up; 5) examining the P for interaction with sex; 6) examining the above associations additionally adjusted for the timing of carbohydrate consumption at dinner and breakfast because the participants had very different eating schedules, i.e., with some participants consuming breakfast at 4:00 a.m. others at 9:00 a.m.; and 7) running energy density models in the energy-adjusted form (carbohydrates intake measured in servings/1,000 kcal/day).

The differences in the studied variables according to the four carbohydrate intake patterns are presented in Table 1. Significant differences in the studied variables were observed for all variables across the four carbohydrate intake patterns (P < 0.001). According to the results, total daily high-quality carbohydrate intake was significantly higher for carbohydrate intake pattern 3 and carbohydrate intake pattern 4 than for the other two carbohydrate intake patterns (P < 0.001), whereas total daily low-quality carbohydrate consumption was significantly higher for carbohydrate intake pattern 2 and carbohydrate intake pattern 4 than for the other two carbohydrate intake patterns (P < 0.001).

The association of total daily quantity of carbohydrates, daily high- and low-quality carbohydrate intake, and carbohydrate intake patterns with all-cause, CVD, and diabetes mortality is shown in Supplementary Table 2. No significant association between total daily quantity of carbohydrates and mortality outcomes was observed. However, compared with the lowest tertile of daily high- and low-quality carbohydrate consumption, the participants who consumed the most high-quality carbohydrates throughout the day (tertile 3) had lower risk of all-cause mortality (HR 0.88; 95% CI 0.79–0.99), whereas the participants who consumed the most low-quality carbohydrates throughout the day (tertile 3) had higher risk of all-cause mortality (HR 1.13; 95% CI 1.01–1.26). Similarly, compared with the participants with carbohydrate intake pattern 1, the participants with carbohydrate intake pattern 3 had a lower risk of all-cause mortality (HR 0.89; 95% CI 0.80–0.99).

The association of Δhigh- and Δlow-quality carbohydrates intake with all-cause, CVD, and diabetes mortality among the four carbohydrate intake patterns is presented in Fig. 2 (HR and 95% CI for tertile 3 vs. tertile 1) and Supplementary Table 3 (complete data). As indicated by HRs and 95% CIs, compared with the participants in the lowest tertile of Δhigh-quality carbohydrate intake, the participants in the highest tertile had a lower risk of all-cause mortality for every carbohydrate intake pattern ([HR 0.82; 95% CI 0.70–0.97] [P for trend = 0.022] for carbohydrate intake pattern 1, [HR 0.79; 95% CI 0.64–0.98] (P for trend = 0.027] for carbohydrate intake pattern 2; [HR 0.83; 95% CI 0.69–1.00] [P for trend = 0.051] for carbohydrate intake pattern 3, and [HR 0.76; 95% CI 0.59–0.98] [P for trend = 0.026] for carbohydrate intake pattern 4), and the participants in the highest tertile with carbohydrate intake pattern 1 also had a lower risk of CVD mortality (HR 0.70; 95% CI 0.52–0.93) (P for trend = 0.017). Moreover, compared with the participants in the lowest tertile of Δlow-quality carbohydrate intake, the participants in the highest tertile had a higher risk for diabetes mortality for carbohydrate intake pattern 3 (HR 1.78; 95% CI 1.02–3.11) (P for trend = 0.056) and carbohydrate intake pattern 4 (HR 3.95; 95% CI 1.28–12.16) (P for trend = 0.024). Furthermore, similar results were still observed in analyzing the daily intake of high- and low-quality carbohydrates at breakfast and dinner. In Supplementary Table 4, compared with the participants in the lowest tertile of low-quality carbohydrate consumption at breakfast, the participants in the highest tertile had greater all-cause mortality risk for carbohydrate intake pattern 4 (HR 1.41; 95% CI 1.07–1.85) (P for trend = 0.025). Additionally, in Supplementary Table 5, compared with the participants in the lowest tertile of low-quality carbohydrate consumption at dinner, the participants in the highest tertile had greater risk of all-cause and diabetes mortality for carbohydrate intake pattern 3 ([HR 1.27; 95% CI 1.03–1.56] [P for trend = 0.020] for all-cause], [HR 2.10; 95% CI 1.11–3.96] [P for trend = 0.033] for diabetes) and carbohydrate intake pattern 4 ([HR 2.77; 95% CI 1.11–6.87] [P for trend = 0.015] for diabetes).

Figure 3 shows the predicted risks of all-cause, CVD, and diabetes mortality in the isocaloric models in switching low-quality carbohydrates to high-quality carbohydrates consumed at dinner among the four carbohydrate intake patterns. Replacing 1 serving low-quality carbohydrates with 1 serving high-quality carbohydrates consumption at dinner was associated with a 19% lower risk of all-cause mortality for carbohydrate intake pattern 1 (HR 0.81; 95% CI 0.74–0.89), a 14% lower risk of all-cause mortality for carbohydrate intake pattern 2 (HR 0.86; 95% CI 0.77–0.96), and a 10% lower risk of all-cause mortality for carbohydrate intake pattern 3 (HR 0.90; 95% CI 0.85–0.95). Furthermore, replacing 1 serving low-quality carbohydrates consumed at dinner with 1 serving high-quality carbohydrates was associated with a 25% lower risk of CVD mortality (HR 0.75; 95% CI 0.64–0.89) for carbohydrate intake pattern 1.

The association of Δhigh- and Δlow-quality carbohydrates intake with mortality was consistently observed in analyzing the association of Δfood intake that was used to calculate the intake of high- and low-quality carbohydrates with mortality (Supplementary Table 6). Higher Δwhole grain intake was related to 18% and 13% lower risks of all-cause mortality for carbohydrate intake pattern 1 (HR 0.82; 95% CI 0.68–0.99) and carbohydrate intake pattern 4 (HR 0.87; 95% CI 0.78–0.97), respectively, and higher Δadded sugar intake was related to a 4% greater diabetes mortality risk for carbohydrate intake pattern 2 (HR 1.04; 95% CI 1.00–1.08). Additionally, higher Δstarchy vegetable intake was related to 26% and 45% greater all-cause (HR 1.26; 95% CI 1.02–1.54) and CVD (HR 1.45; 95% CI 1.03–2.04) mortality risks for carbohydrate intake pattern 3. Furthermore, in excluding participants who skipped breakfast or dinner or whose follow-up time was <2 years, the association in these sensitivity analyses was consistent with the results of the total sample (Supplementary Tables 7–9). Furthermore, sex did not have modification effects on the above association (Supplementary Table 10). The above association remained robust when additional adjustment was made for the timing of breakfast and dinner consumption or using the energy-adjusted form of total daily carbohydrate intake (Supplementary Tables 11 and 12).

In this study we examined the association of quantity, quality, and timing of carbohydrate intake with all-cause, CVD, and diabetes mortality. We observed that consuming more high-quality carbohydrates throughout the day was associated with lower all-cause mortality, whereas more daily intake of low-quality carbohydrates was associated with greater all-cause mortality. More importantly, consuming more high-quality carbohydrates at dinner was related to lower CVD and all-cause mortality risk, and consuming more low-quality carbohydrates at dinner was related to greater diabetes mortality irrespective of the total daily quantity and quality of carbohydrates. Regarding the exact food groups, consuming more whole grains at dinner reduced total mortality, whereas consuming more starchy vegetables and added sugars at dinner was associated with increased all-cause, CVD, and diabetes mortality. Furthermore, isocaloric replacement of low-quality carbohydrates consumption at dinner with high-quality carbohydrates could reduce the risk of all-cause and CVD mortality among the participants whose total daily high- and low-quality carbohydrate consumption amounts were both below the median.

In this study, no significant association between total daily intake quantity of carbohydrates and mortality outcomes was observed, whereas a significant association between total daily high- or low-quality carbohydrate intake and decreased or increased risk of all-cause mortality was observed, suggesting that the quality of daily carbohydrate intake was more important than the quantity. These findings were in line with those of previous studies, which all emphasized the importance of carbohydrate quality in the prevention of CVD and improving life expectancy (3,19–22). More importantly, in this study we found that in addition to the quantity and quality of daily carbohydrates consumed, the timing of carbohydrate consumption also plays an important role in reducing CVD, diabetes, and all-cause mortality. This study showed that among the participants whose total daily high- and low-quality carbohydrate consumption was at low levels, consuming more high-quality carbohydrates at dinner was associated with lower risks of CVD and all-cause mortality, which suggested that participants with a relatively low total daily carbohydrate intake could consume high-quality carbohydrates at dinner to decrease the risks of CVD and all-cause mortality. Furthermore, we also observed that among the participants whose daily amount of high-quality carbohydrates consumed was high and low-quality carbohydrates consumed was low, consuming more low-quality carbohydrates at dinner was associated with greater diabetes mortality. This observation suggested that although participants had a relatively high daily total intake of high-quality carbohydrates and reduced their daily total intake of low-quality carbohydrates, they should also focus on their timing of low-quality carbohydrate consumption to decrease diabetes mortality risk. Moreover, this association was relatively robust and was independent of a series of dietary covariates, including skipping breakfast, skipping dinner, daily energy intake, and dietary quality.

The above observation can be partially supported by previous studies. Recent studies documented that the distribution of energy and macronutrient intake across meals throughout the day was associated with the prevalence of cardiometabolic disease and CVD and all-cause mortality (13,23). Moreover, findings of a few randomized controlled clinical trials showed that high levels of carbohydrate consumption at night could increase postprandial glucose the next morning (24,25), and compared with the same level of food consumption in the morning, late-night consumption of higher glycemic load foods could increase glucose and insulin levels, suggesting that a higher–glycemic index meal in the evening would have a more substantial influence on glucose and insulin levels (26). Additionally, the findings of this study show that replacing low-quality carbohydrates intake at dinner with high-quality carbohydrates significantly decreased the risk of total and CVD mortality among the participants whose daily total carbohydrate intake was low. High-quality carbohydrates are rich in fiber, vitamins, and phytochemicals, such as flavonoids (7,27). Previous studies have shown that gut microbes and their metabolites have a circadian pattern, and the bacteria that produce short-chain fatty acids using dietary fiber or phytochemicals are usually more active in the evening (28,29). Therefore, consuming more high-quality carbohydrates at dinner may provide more fuel for bacteria to produce more short-chain fatty acids. Investigators of a recent study also found that vegetable-based dinners rich in dietary fiber lowered the risks of CVD, cancer, and all-cause mortality (30). Taken together, the evidence from this study and previous studies indicates that the timing of carbohydrate consumption is as important as the total quantity and quality of the carbohydrates.

Additionally, the findings of this study also reveal that consuming more whole grains at dinner was associated with lower all-cause mortality and consuming more starchy vegetables or added sugars at dinner was associated with greater risks of CVD, diabetes, and all-cause mortality. The metabolism of glucose and insulin has a diurnal variation, and metabolism is frequently impaired at night (31). In a previous study findings showed that postprandial glucose and lipid tolerance were relatively impaired when meals were consumed at night compared with when the same foods were consumed during the daytime (32), and participants who consumed starchy snacks after dinner were more likely to die due to CVD and all causes (30), which may further support the findings in this study. In summary, the results of this study illustrate that the timing of carbohydrate consumption plays an important role in health equal to that of the overall quality of carbohydrates consumed throughout the day, and these are both more important than the overall quantity of daily total carbohydrates.

This study has a number of strengths. First, we demonstrate the association of carbohydrate intake with CVD mortality, diabetes mortality, and total mortality across different carbohydrate intake patterns by simultaneously considering the overall quantity, quality, and timing of carbohydrate intake based on a well-designed study (NHANES). Second, the association was proven to be robust across all four groups, which showed that the conclusions were independent of the total consumption of high-quality and low-quality carbohydrates throughout the day. Third, this study involved a large number of U.S. adults, with a prospective study design. Fourth, the association was still robust when the factors of skipping breakfast, skipping dinner, timing of breakfast and dinner, timing of follow-up, and energy-adjusted form were taken into consideration, and these factors are classic dietary confounders. However, this study still has limitations. First, dietary information used was gathered at a single baseline visit, and the participants might have changed eating habits during the follow-up period. Second, causality could not be determined due to the observational study design. Third, the current conclusion is applicable only to U.S. adults, and the situations of other ethnic groups still need to be verified. Finally, although a number of confounders were included, unmeasured confounders are still likely.

This study indicates that consuming more low-quality carbohydrates at dinner was associated with greater diabetes mortality risk, whereas consuming more high-quality carbohydrates at dinner can lower the risk of total and CVD mortality irrespective of the total daily quantity and quality of the carbohydrates. This study illustrates that the timing of carbohydrate consumption plays an important role in health equal to that of the overall quality of carbohydrates consumed throughout the day.

Acknowledgments. The authors thank the participants and staff of NHANES 2003–2014 for their valuable contributions.

Funding. This study was supported by funds from the National Natural Science Foundation of China (82073534 to C.S.) and HMU Marshal Initiative Funding (HMUMIF-21010 to T.H.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. All authors contributed substantially to this study. W.J. and C.S. designed the study and served as guarantors, taking responsibility for the data’s integrity and the data analysis’s accuracy. W.H. and T.H. wrote the manuscript. X.S. and Y.C. conducted the statistical analysis. J.X., Y.W., and X.Y. critically revised the manuscript. All authors critically reviewed the manuscript and approved the final version for publication. W.J. and C.S. were responsible for the overall manuscript.