Content uploaded by Gang Liu
Author content
All content in this area was uploaded by Gang Liu on Sep 12, 2022
Content may be subject to copyright.
The Journal of Clinical Endocrinology & Metabolism, 2022, 107, e2702–e2709
https://doi.org/10.1210/clinem/dgac235
Advance access publication 16 April 2022
Clinical Research Article
Received: 30 November 2021. Editorial Decision: 11 April 2022. Corrected and Typeset: 6 May 2022
© The Author(s) 2022. Published by Oxford University Press on behalf of the Endocrine Society. All rights reserved. For permissions, please e-mail:
journals.permissions@oup.com
Associations of Moderate Low-Carbohydrate Diets
With Mortality Among Patients With Type 2 Diabetes:
AProspective CohortStudy
ZhenzhenWan,1,* ZhileiShan,1,* TingtingGeng,1,2 QiLu,1, LinLi,1 JiaweiYin,1 LiegangLiu,1
AnPan,2, and GangLiu1,
1Department of Nutrition and Food Hygiene, Hubei Key Laboratory of Food Nutrition and Safety, Ministry of Education Key Lab of Environment
and Health, and State Key Laboratory of Environmental Health (Incubating), School of Public Health, Tongji Medical College, Huazhong
University of Science and Technology, 430030 Wuhan, China
2Department of Epidemiology and Biostatistics, School of Public Health, Tongji Medical College, Huazhong University of Science and
Technology, 430030 Wuhan, China
*Z.W.and Z.S.contributed equally to this work as co-first authors
Correspondence: Gang Liu, PhD, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, 13 Hangkong Rd, 430030
Wuhan, China. Email: liugang026@hust.edu.cn.
Abstract
Context: A statement of context for the abstract was added in Objective as follows: Lower-carbohydrate-diet (LCD) has been reported to have
beneficial effects on cardiovascular risk factor profile in general population. However, whether adherence to an LCD could benefit long-term
survival among individuals with diabetes is unclear.
Objective: This work aimed to investigate the associations of different types of lower-carbohydrate diets with mortality among individuals with
type 2 diabetes (T2D).
Methods: This prospective study included 5677 patients with T2D. The overall, unhealthy, and healthy lower-carbohydrate-diet (LCD) scores
were calculated based on the percentage of energy from total and subtypes of carbohydrate, protein, and fat. Deaths were determined via
linkage to the National Death Index records until December 31, 2015. Cox proportional hazards models were used to estimate the hazard ratios
(HRs) and 95% CIs of mortality.
Results: During a median of 6.3 years of follow-up (39 401 person-years), 1432 deaths were documented. After multivariable adjustment
including lifestyle factors, diabetes duration, and glycated hemoglobin A1c, patients in the third quartile of overall LCD score had the lowest risk of
mortality (HR: 0.65; 95% CI, 0.50-0.85), compared with the first quartile. The multivariable-adjusted HRs (95% CIs) of mortality across quartiles
of healthy lower-carbohydrate-diet score were 1.00 (reference), 0.78 (0.64-0.96), 0.73 (0.58-0.91), and 0.74 (0.58-0.95) (Ptrend = .01). Isocalorically
replacing 2% of energy from carbohydrates with plant-based protein or polyunsaturated fatty acids was associated with 23% to approximately
37% lower total mortality. Similar results were observed when analyses were stratified by age, sex, race/ethnicity, smoking status, body mass
index, physical activity, and diabetes duration.
Conclusion: Healthy LCD score was significantly associated with a lower risk of mortality in adults with T2D. Adherence to a well-balanced
moderate lower-carbohydrate diet that emphasizes healthy carbohydrates, plant-based protein, and polyunsaturated fat may prevent premature
death among patients with T2D.
Key Words: lower-carbohydrate-diet, prospective study, type 2 diabetes, mortality
Abbreviations: BMI, body mass index; CRP, C-reactive protein; CVD, cardiovascular disease; HbA1c, glycated hemoglobin A1c; HDL, high-density lipoprotein;
HOMA-IR, homeostasis model assessment of insulin resistance; HR, hazard ratio; LCD, lower-carbohydrate diet; LDL, low-density lipoprotein; NHANES, National
Health and Nutrition Examination Survey; T2D, type 2 diabetes.
Diabetes is a serious public health issue associated with
high morbidity and mortality rates (1). According to the
International Diabetes Federation Diabetes Atlas Report in
2019, approximately 4.2 million deaths among adults aged
20 to 79 years are attributable to diabetes (2). Among the
modiable risk factors, healthy diet has played an essential
role in preventing and improving complications of diabetes
(3, 4).
A lower-carbohydrate diet (LCD), dened as reduced
carbohydrate and increased fat and protein contributions to
total energy, has been suggested to exert favorable effects on
weight loss, glycemic control, and reduced glycated hemo-
globin A1c (HbA1c), even beyond the energy restriction (5-10).
LCD has also been reported to have benecial effects on car-
diovascular risk factor prole, such as circulating cholesterol
and triglycerides (11). However, the association between ad-
herence to LCD and risk of mortality among general popu-
lations is less conclusive, with some studies showing that a
higher LCD score was associated with a higher risk of mor-
tality (12-14), whereas other studies reported an inverse as-
sociation (15) or null association (16, 17). Moreover, besides
the quantity, different quality and sources of carbohydrate,
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022
The Journal of Clinical Endocrinology & Metabolism, 2022, Vol. 107, No. 7 e2703
fat, and protein could yield contrasting health effects (18, 19).
Several studies indicated that an LCD highlighting animal
sources of protein and fat was associated with a higher risk
of all-cause mortality and cardiovascular disease (CVD) mor-
tality, whereas an LCD rich in plant-based foods was associ-
ated with a lower risk of mortality (13, 20, 21).
Although potential benets of LCD are highlighted by a
US Consensus Report for glycemic control among patients
with T2D (22), whether adherence to an LCD could benet
long-term survival among individuals with diabetes is unclear.
Furthermore, whether the LCD-mortality association among
diabetes would vary by sex, race/ethnicity, obese status, or dia-
betes duration remains unknown. To address these research
gaps, for the rst time, we prospectively investigated the as-
sociations of different LCD scores with mortality among pa-
tients with T2D, using data from the National Health and
Nutrition Examination Survey (NHANES). We hypothesized
that an LCD score that emphasized healthy carbohydrates,
plant-based protein, and unsaturated fat would be associated
with a lower risk of mortality.
Materials andMethods
Study Population
The NHANES is a periodic survey conducted by the National
Center for Health Statistics of the Centers for Disease
Control, which represents a nationally representative sample
of the noninstitutionalized US civilian population. In this pro-
spective study, we included participants aged 20years and
older with diabetes who had at least one reliable dietary re-
call data from 8cycles of NHANES between 1999 and 2014.
Those with gestational diabetes at baseline were excluded.
The details of sampling method and analytic guidelines have
been published elsewhere (23).
Diabetes was dened as self-reported physical diagnosis of
diabetes, use of oral glucose-lowering medicines or insulin,
fasting glucose greater than or equal to 7.0 mmol/L, per-
centage HbA1c greater than or equal to 6.5%, or oral glucose
tolerance test greater than or equal to 11.1mmol/L. After ex-
clusion of individuals with unreliable energy intake (< 800
or > 4200 kcal/day for men and < 600 or > 3500 kcal/day for
women), with incomplete information on mortality, or who
were self-reported pregnant at baseline, 5677 patients with
diabetes were included in the nal analysis. Aowchart of
the study participants is shown in Supplementary Fig. 1 (24).
NHANES study protocols were approved by the institutional
review board of the National Center of Health Statistics.
Dietary Assessment and Lower-Carbohydrate Diet
Score Computation
Diet intake was assessed using 1 24-hour dietary recall in
NHANES 1999 to 2000, and 2 24-hour dietary recalls in
2001 to 2014; means of 2 values for macronutrients were
used. We also applied the National Cancer Institute method
in estimation of usual intake of macronutrients to reduce
measurement error (25). Three LCD scores, that is, overall
LCD, unhealthy LCD, and healthy LCD, were computed
using the method described in our previous study (20). Briey,
we dened carbohydrates from whole grains, whole fruit, leg-
umes, and nonstarchy vegetables as high-quality carbohy-
drates, and carbohydrates from rened grains, added sugar,
fruit juice, potato, other starchy vegetables, and other sources
as low-quality carbohydrates. We used the percentage of en-
ergy from fat, protein, and carbohydrates to represent dietary
composition instead of the absolute intake to minimize the
underreporting bias (13). The participants were divided into
11 strata (0-10) for the percentage of energy from fat (satur-
ated or unsaturated), protein (animal-based or plant-based),
and carbohydrate (high-quality or low-quality). Individuals
at higher categories of fat and protein components were as-
signed higher scores whereas carbohydrate component was
reversely scored (Supplementary Table 1) (24). The overall
LCD score was the sum of the 3 macronutrients compo-
nent scores, ranging from 0 to 30. An unhealthy LCD score
was computed according to the percentage of energy from
high-quality carbohydrates, animal-based protein, and satur-
ated fat; and a healthy LCD was computed based on low-
quality carbohydrates, plant-based protein, and unsaturated
fat (20). The higher the score, the more closely the participant
follow a healthy diet pattern.
Assessment of Covariates
Information on age, sex, race/ethnicity, education level, family
income, smoking status, physical activity, diabetes, and med-
ical history was collected at recruitment by trained inter-
viewers using standardized questionnaires. Height and body
weight were obtained through physical examinations. Body
mass index (BMI) was calculated as weight in kilograms
divided by height in meters squared, and then classied as
less than 25, 25 to 30, and 30 or greater. Physical activity
in leisure time was calculated by summarizing times of self-
reported moderate to vigorous activity per week. Alcohol in-
take data were collected at a mobile examinationcenter.
In addition, plasma glucose, insulin, HbA1c, triglycerides,
total cholesterol, high-density lipoprotein (HDL) cholesterol,
low-density lipoprotein (LDL) cholesterol, and C-reactive
protein (CRP) were measured at recruitment. We computed
the homeostasis model assessment of insulin resistance
(HOMA-IR) using the method of Matthews etal (26).
Ascertainment of Mortality
Mortality was determined using the NHANES Public-Use
Linked Mortality File as of December 31, 2015. This le
linked NHANES to the National Death Index through a
rigorous probability matching and death certicate review
process. All-cause mortality consisted of all specied and un-
known causes.
Statistical Analysis
Considering the complex sampling design of NHANES, all
analyses in the present study incorporated sample weights,
clustering, and stratication. Person-time was computed from
the date of the dietary interview to the date of death or the
end of follow-up (December 31, 2015), whichever came rst.
Percentages of missing values were less than 5%, except for
family income-poverty ratio (8.5%). Multiple imputation was
performed for missing covariates values. SAS PROC MI and
PROC MIANALYZE were used to maximize data availability
for all variables. We imputed 5 data sets to achieve less-biased
and robust results over different simulations.
The generalized linear model was used to examine the as-
sociations of LCD scores with cardiometabolic biomarkers
at baseline, including plasma glucose, insulin, HOMA-IR,
HbA1c, triglycerides, CRP, total cholesterol, HDL, and LDL.
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022
e2704 The Journal of Clinical Endocrinology & Metabolism, 2022, Vol. 107, No. 7
Cox proportional hazards regression model was used to in-
vestigate the associations between LCD scores and mortality
with the lowest quartile as the reference group. Proportional
hazards assumptions were tested using Schoenfeld residuals
method and no violations were observed for all models. In
model 1, we adjusted for age (continuous, years), sex (male,
female), and race (non-Hispanic White, non-Hispanic Black,
Mexican American, or other). In model 2, we further adjusted
for total energy intake (continuous, kcal/day), cholesterol in-
take (in quartiles, mg), educational level (< high school, high
school or equivalent, or ≥ college), family income-poverty
ratio (0-1.0, 1.0-3.0, or > 3.0), BMI (< 25, 25-30, or ≥ 30),
smoking status (never, former, or current smoker), alcohol
drinking (nondrinker, low to moderate drinker, or heavy
drinker), leisure-time moderate-to-vigorous physical activity
(0, 1-3, or > 3 times/week), duration of diabetes (≤ 3, 3-10,
or > 10years), diabetes medication use (none, only oral medi-
cation, insulin, or others), HbA1c (< 7.0% or ≥ 7.0%), and
history of hypertension, hypercholesterolemia, or CVD (yes
or no). Categorization of LCD scores was set as the primary
analysis. To test for linear trend, we modeled LCD scores by
assigning the median value to each quartile. As a secondary
analysis, a 20% increment in LCD scores was also applied
to estimate the hazard ratios (HRs) and 95% CIs of the risk
of all-cause mortality. Isocaloric models were used to esti-
mate HRs of mortality when energy from total, unhealthy or
healthy carbohydrates was theoretically replaced by equiva-
lent energy from animal or plant-based protein, or different
types of fat (27). Results for 2% energy substitution were
shown to make the HRs comparable for various macronu-
trient substitution analyses (28).
We further stratied the analyses by age at recruitment (< 65
or ≥ 65years), sex (male or female), race/ethnicity (White or
non-White), smoking status (never or ever smokers), drinking
status (nondrinkers or drinkers), physically activity (yes or
no), BMI (< 30 or ≥ 30), diabetes duration (≤ 3 or > 3years),
and history of comorbidities (yes or no). Both categorical and
continuous (per 20% increment) LCD scores were used to es-
timate the HRs and 95% CIs of the risk of all-cause mortality.
The P values for the product terms between LCD scores and
stratication variables were used to assess the signicance of
interactions.
A series of sensitivity analyses were also conducted to test
the robustness of the results. To minimize the potential reverse
causality, we excluded participants who died within the rst
year of follow-up. Second, we further excluded participants
with a history of CVD or cancer. Third, to explore the poten-
tial mediation effects of lipid and inammation, we further
adjusted for HDL and CRP (with available data in NHANES
1999-2010). Finally, to address the potential sex-difference
issue, we generated the LCD scores using sex-specic macro-
nutrient components and repeated the main analyses. SAS
statistical software (version 9.4, SAS Institute Inc) was used
for all analyses, and a 2-tailed P value less than .05 was con-
sidered statistically signicant.
Results
Among the 5677 patients with T2D, the mean (SD) age was
61.8 (13.5) years and 49.7% were female. The baseline
characteristics of the participants according to quartiles of
LCD scores are shown in Table 1. Participants in the highest
quartiles of overall, unhealthy, or healthy LCD scores con-
suming 45.9% to 47.4% of energy from carbohydrate intake.
Participants who had a higher overall LCD score tended to
be younger, male, non-Hispanic White, and physically active,
and had higher family income and education level, higher in-
take of cholesterol, and lower intake of total energy. Similar
results were observed for the other 2 LCD scores, with the
exception that participants with a higher healthy LCD score
were older and had a better diet pattern while participants
with a higher unhealthy LCD score tended to be physically
inactive and had a higher intake of totalenergy.
The least-square means of cardiometabolic biomarkers ac-
cording to the healthy LCD score are shown in the Table 2.
Higher healthy LCD score was signicantly associated with
lower levels of CRP and higher levels of HDL at baseline (all
Ptrend ≤ .01). In addition, higher overall LCD score was as-
sociated with lower triglycerides and higher HDL, whereas
unhealthy LCD score was associated with higher HDL
(Supplementary Tables 2 and 3)(24).
During a median of 6.3years of follow-up (39 401 person-
years), we documented 1432deaths. Compared with the rst
quartile, participants in the third quartile of overall LCD
score had the lowest risk of all-cause mortality, with an HR
(95% CI) of 0.65 (0.50-0.85). The multivariable-adjusted
HRs (95% CIs) of all-cause mortality across quartiles of
healthy LCD score were 1.00 (reference), 0.78 (0.64-0.96),
0.73 (0.58-0.91), and 0.74 (0.58-0.95) (Ptrend = .01) (Table 3).
For 20% increments in the healthy LCD score, the HR (95%
CI) of all-cause mortality was 0.89 (0.81-0.97) (Table 3). No
signicant association was observed between unhealthy LCD
scores and mortality. The results remained largely unchanged
when BMI, HbA1c, and diabetes duration were included as
continuous variables into the models. Isocalorically replacing
2% of energy from total carbohydrates, unhealthy or healthy
carbohydrates with plant-based protein or PUFAs was associ-
ated with 23% to approximately 37% lower total mortality
(Fig. 1).
Results of stratied and sensitive analyses are shown in
the online repository (24). Similar results were demonstrated
when analyses were stratied by age, sex, race/ethnicity,
smoking and drinking status, physical activity, BMI, diabetes
duration, and presence of comorbidity, although some of the
associations did not reach statistical signicance largely be-
cause of reduced power. We did not observe any signicant
interactions between LCD scores and the stratied factors
on all-cause mortality after correcting for multiple testing
(Supplementary Figs. 2 and 3)(24).
In sensitive analyses, the association between healthy LCD
and all-cause mortality remained signicant when excluding
participants who died during the rst year of follow-up, or
those with CVD or cancer at baseline (Supplementary Table
4) (24). The results were slightly attenuated when further
adjusting for HDL and CRP (Supplementary Table 5) (24).
Similar ndings were observed when using sex-specic LCDs
(Supplementary Table 6)(24).
Discussion
In this large, prospective study of patients with T2D, a healthy
LCD score was associated with a lower risk of mortality, in-
dependent of lifestyle factors, diabetes duration, and glucose
control. Replacing carbohydrates with PUFAs or plant-based
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022
The Journal of Clinical Endocrinology & Metabolism, 2022, Vol. 107, No. 7 e2705
Table 1. Characteristics of participants with diabetes according to quartiles of lower-carbohydrate-diet scores in the National Health and Nutrition
Examination Survey 1999 to 2014a
Characteristics Overall LCD score Unhealthy LCD score Healthy LCD score
Quartile 1 Quartile 4 Quartile 1 Quartile 4 Quartile 1 Quartile 4
Participants, No. 1,490 1,544 1,487 1,497 1,301 1,345
Median score (IQR) 6 (4-8) 24 (22-26) 7 (5-9) 23 (21-25) 6 (4-8) 23 (22-25)
Age, mean (SD), y 61.7 (14.7) 60.7 (13.1) 65.2 (12.3) 57.9 (13.9) 59.0 (15.3) 63.8 (11.7)
Age at diagnosis, mean (SD), y 50.0 (16.5) 48.6 (15.6) 52.4 (15.4) 47.1 (15.5) 48.3 (16.6) 50.8 (15.4)
BMI, mean (SD) 32.0 (7.4) 32.6 (7.5) 31.1 (6.8) 33.1 (7.8) 32.7 (7.6) 31.4 (7.0)
Female 1015 (68.1) 515 (33.4) 1013 (68.1) 469 (31.3) 728 (56.0) 566 (42.1)
Race/ethnicity
Non-Hispanic White 468 (31.4) 753 (48.8) 430 (28.9) 784 (52.4) 465 (35.7) 572 (42.5)
Non-Hispanic Black 445 (29.9) 359 (23.3) 362 (24.3) 396 (26.5) 431 (33.1) 281 (20.9)
Hispanic 331 (22.2) 278 (18.0) 396 (26.6) 206 (13.8) 244 (18.8) 286 (21.3)
Other 246 (16.5) 154 (10.0) 299 (20.1) 111 (7.4) 161 (12.4) 206 (15.3)
Family income to poverty ratio
≤ 1.0 356 (26.4) 298 (21.1) 325 (24.1) 346 (25.0) 354 (27.2) 221 (16.4)
1.0-3.0 663 (49.1) 597 (42.2) 665 (49.4) 605 (43.7) 698 (53.7) 664 (49.4)
> 3.0 332 (24.6) 519 (36.7) 357 (26.5) 434 (31.3) 249 (19.1) 460 (34.2)
Educational level
< High school 614 (41.4) 558 (36.1) 610 (41.1) 555 (37.1) 538 (41.4) 470 (34.9)
High school or equivalent 340 (22.9) 362 (23.5) 327 (22.1) 366 (24.5) 317 (24.4) 335 (24.9)
≥ College 531 (35.8) 624 (40.4) 546 (36.8) 576 (38.5) 446 (34.3) 540 (40.2)
Nonsmoker 832 (56.0) 640 (41.5) 896 (60.3) 593 (39.6) 624 (48.0) 607 (45.1)
Nondrinker 657 (46.3) 432 (29.2) 709 (50.3) 402 (28.1) 494 (38.0) 442 (32.9)
Physical activity, times/wk
0 995 (66.8) 967 (62.6) 914 (61.5) 976 (65.2) 890 (68.4) 789 (58.7)
1-2 150 (10.1) 171 (11.1) 143 (9.6) 175 (11.7) 151 (11.6) 149 (11.1)
≥3 345 (23.2) 406 (26.3) 430 (28.9) 346 (23.1) 260 (20.0) 407 (30.3)
Duration of diabetes, y
≤ 3 751 (51.6) 622 (41.3) 662 (45.5) 719 (49.1) 696 (53.5) 465 (34.6)
3-10 306 (21.0) 386 (25.7) 326 (22.4) 342 (23.4) 311 (23.9) 410 (30.5)
> 10 399 (27.4) 497 (33.0) 466 (32.1) 403 (27.5) 294 (22.6) 470 (34.9)
Dietary intake, mean (SD)
Total energy, kcal/d 1866 (401) 1829 (372) 1780 (373) 1905 (396) 1937 (410) 1802 (368)
Total carbohydrate, % of total energy intake 56.8 (2.5) 45.9 (2.6) 55.4 (3.5) 47.0 (3.5) 55.7 (3.2) 47.4 (3.6)
High-quality carbohydrate 10.8 (4.4) 8.6 (3.3) 13.7 (3.8) 6.7 (2.1) 8.1 (3.1) 11.4 (4.3)
Low-quality carbohydrate 46.0 (4.7) 37.4 (3.5) 41.7 (5.6) 40.3 (4.2) 47.7 (3.4) 36.0 (3.1)
Total protein, % of total energy intake 15.3 (1.4) 17.9 (1.4) 16.0 (1.6) 17.4 (1.6) 15.4 (1.5) 17.6 (1.6)
Animal protein 9.6 (1.1) 12.0 (1.4) 9.7 (1.2) 11.9 (1.4) 10.2 (1.4) 11.1 (1.6)
Plant protein 5.7 (0.8) 5.9 (0.8) 6.3 (0.9) 5.5 (0.7) 5.2 (0.5) 6.5 (0.7)
Total fat, % of total energy intake 27.9 (2.7) 36.2 (3.0) 28.6 (3.4) 35.6 (3.3) 28.9 (3.0) 35.0 (3.6)
Saturated fat 9.8 (1.4) 12.8 (1.6) 9.6 (1.2) 13.1 (1.4) 10.5 (1.6) 11.9 (1.8)
PUFAs 7.1 (1.0) 8.7 (1.2) 7.6 (1.2) 8.2 (1.3) 6.9 (0.9) 8.9 (1.2)
MUFAs 11.1 (1.3) 14.7 (1.5) 11.4 (1.6) 14.3 (1.7) 11.4 (1.3) 14.3 (1.8)
Total cholesterol intake, mg/d (IQR) 163 (85-275) 302 (175-501) 142 (78-242) 315 (181-514) 199 (112-342) 232 (135-419)
HbA1c, ≥ 7.0% 601 (41.6) 691 (46.3) 594 (41.5) 645 (45.0) 532 (40.9) 573 (42.6)
History of CVD 380 (25.5) 391 (25.3) 413 (27.8) 388 (25.9) 332 (25.5) 332 (24.7)
Hypertension 953 (64.0) 1,012 (65.5) 1,001 (67.3) 939 (62.7) 789 (60.7) 900 (66.9)
Hypercholesterolemia 765 (51.3) 841 (54.5) 800 (53.8) 778 (52.0) 618 (47.5) 771 (57.3)
Abbreviations: BMI, body mass index; CVD, cardiovascular disease; HbA1c, glycated hemoglobin A1c; IQR, interquartile range; LCD, lower-carbohydrate
diet; MUFAs, monounsaturated fatty acids; PUFAs, polyunsaturated fatty acids.
aData are presented as means (SD) for continuous variables and numbers (%) for categorical variables.
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022
e2706 The Journal of Clinical Endocrinology & Metabolism, 2022, Vol. 107, No. 7
protein was also associated with lower mortality. The ndings
were robust in a series of sensitivity analyses and stratied
analyses. In addition, higher healthy LCD score was signi-
cantly associated with lower level of CRP and higher level
of HDL. No association was observed between overall and
unhealthy LCD scores and mortality.
Although an LCD has become popular in rapid weight loss
with low carbohydrates and high protein and fat intake (7,
Table 2. Least-square means of cardiometabolic markers according to healthy lower-carbohydrate diet score among participants with diabetes in the
National Health and Nutrition Examination Survey 1999 to 2014
Healthy LCD score
Ptrend Quartile 1 (0-9) Quartile 2 (10-15) Quartile 3 (16-20) Quartile 4 (21-30)
Glucose (n = 2974, mmol/L) 9.15 ± 0.17 9.15 ± 0.16 8.98 ± 0.16 9.01 ± 0.16 .25
Insulin (n = 2904, μU/mL) 24.8 ± 1.72 21.9 ± 1.63 23.7 ± 1.66 25.2 ± 1.67 .65
HOMA-IR (n = 2895) 10.2 ± 0.72 9.35 ± 0.68 9.98 ± 0.70 10.7 ± 0.70 .42
HbA1c (n = 5489, %) 7.65 ± 0.07 7.61 ± 0.07 7.58 ± 0.07 7.53 ± 0.07 .09
Triglycerides (n = 2944, mmol/L) 2.11 ± 0.12 2.06 ± 0.11 1.95 ± 0.11 2.00 ± 0.11 .23
CRP (n = 3922, mg/dL) 0.70 ± 0.06 0.65 ± 0.06 0.55 ± 0.06 0.58 ± 0.06 .01
Total cholesterol (n = 5396, mmol/L) 5.05 ± 0.05 5.00 ± 0.05 4.95 ± 0.05 4.98 ± 0.05 .10
HDL (n = 5395, mmol/L) 1.32 ± 0.01 1.36 ± 0.01 1.36 ± 0.01 1.38 ± 0.01 < .001
LDL (n = 2680, mmol/L) 2.78 ± 0.06 2.77 ± 0.05 2.69 ± 0.06 2.73 ± 0.06 .18
The least squares (mean ± SE) were estimated using the general linear model with the adjustment for age (continuous), sex (male, or female), race/ethnicity
(non-Hispanic White, non-Hispanic Black, Mexican American, or other), total energy (continuous), BMI (< 25, 25-30, or ≥ 30), education level (< high
school, high school or equivalent, or ≥ college), family income-poverty ratio (0-1.0, 1.0-3.0, or > 3.0), drinking status (nondrinker, low-to-moderate
drinker, or heavy drinker), smoking status (never smoker, former smoker, or current smoker), and leisure-time moderate-to-vigorous physical activity (0,
1-3, or > 3 times/week), duration of diabetes (< 3, 3-10, or ≥ 10years), diabetes medication use (none, only oral medication, insulin, or others), HbA1c
(< 7.0% or ≥ 7.0%), cholesterol intake (in quartiles), self-reported hypertension (yes or no), hypercholesterolemia (yes or no), CVD (yes or no), and
cancer (yes or no).
Abbreviations: BMI, body mass index; CRP, C-reactive protein; CVD, cardiovascular disease; HbA1c, glycated hemoglobin A1c; HDL, high-density
lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; LCD, lower-carbohydrate diet; LDL, low-density lipoprotein.
Table 3. Associations between lower-carbohydrate diet scores and all-cause mortality among participants with diabetes in the National Health and
Nutrition Examination Survey 1999 to 2014
Quartiles of LCD scores
Ptrend
Per 20% increment
in score Quartile 1 Quartile 2 Quartile 3 Quartile 4
Overall LCD score
Median score (IQR) 6 (4-8) 12 (11-13) 17 (16-19) 24 (22-26)
Person-years of follow-up 10 656 8351 9855 10 539
Deaths, No. 390 318 353 371
Model 1a1.00 0.74 (0.58-0.95) 0.67 (0.53-0.85) 0.81 (0.62-1.06) .15 0.92 (0.85-1.00)
Model 2b1.00 0.76 (0.59-0.98) 0.65 (0.50-0.85) 0.81 (0.60-1.09) .16 0.91 (0.84-1.00)
Unhealthy LCD score
Median score (IQR) 7 (5-9) 13 (12-14) 17 (16-18) 23 (21-25)
Person-y of follow-up 10 305 8240 10 464 10 392
Deaths, No. 391 318 379 344
Model 1a1.00 1.17 (0.95-1.43) 1.07 (0.88-1.30) 1.09 (0.86-1.37) .60 1.01 (0.94-1.10)
Model 2b1.00 1.24 (0.99-1.56) 1.08 (0.88-1.34) 1.02 (0.79-1.31) .92 0.98 (0.90-1.07)
Healthy LCD score
Median score (IQR) 6 (4-8) 13 (11-14) 18 (17-19) 23 (22-25)
Person-years of follow-up 9415 11 042 9935 9009
Deaths, No. 318 426 388 300
Model 1a1.00 0.79 (0.64-0.97) 0.71 (0.57-0.90) 0.69 (0.54-0.88) .002 0.87 (0.80-0.95)
Model 2b1.00 0.78 (0.64-0.96) 0.73 (0.58-0.91) 0.74 (0.58-0.95) .02 0.89 (0.81-0.97)
Abbreviations: BMI, body mass index; CVD, cardiovascular disease; HbA1c, glycated hemoglobin A1c; IQR, interquartile range; LCD, lower-carbohydrate
diet.
aModel 1: adjusted for age (continuous), sex (male, or female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, or other).
bModel 2: further adjusted for total energy (continuous), BMI (< 25, 25-30, or ≥ 30), education level (< high school, high school or equivalent, or ≥ college),
family income-poverty ratio (0-1.0, 1.0-3.0, or > 3.0), drinking status (nondrinker, low-to-moderate drinker, or heavy drinker), smoking status (never
smoker, former smoker, or current smoker), and leisure-time moderate-to-vigorous physical activity (0, 1-3, or > 3 times/week), duration of diabetes (<3,
3-10, or ≥ 10years), diabetes medication use (none, only oral medication, insulin, or others), HbA1c (< 7.0% or ≥ 7.0%), cholesterol intake (in quartiles),
self-reported hypertension (yes or no), hypercholesterolemia (yes or no), CVD (yes or no), and cancer (yes or no).
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022
The Journal of Clinical Endocrinology & Metabolism, 2022, Vol. 107, No. 7 e2707
29, 30), the associations between LCD and risk of CVD and
long-term survival among general populations were incon-
sistent. For instance, studies using data from NHANES and The
Scandinavian Women’s Lifestyle and Health Cohort showed
that LCD was associated with higher total and CVD mor-
tality (12, 31), whereas the Nurses’ Health Study and a cohort
study in Sweden observed null association of LCD with CVD
events or mortality (16, 17). In addition, the Prospective Urban
Rural Epidemiology (PURE) study (including 135 335 individ-
uals) and a cohort study in Japanese individuals found that
lower carbohydrate intake was associated with lower mortality
(15, 32). Moreover, the Atherosclerosis Risk in Communities
(ARIC) study including 15 428 participants with 25 years of
follow-up suggested a U-shaped association between carbo-
hydrate intake and mortality with a nadir at consuming 50%
to 55% energy from carbohydrates (33). The heterogeneity of
these studies may be partially due to overlooking the sources
of macronutrients: animal-based or plant-based. Increasing
evidence has indicated that adherence to an LCD that empha-
sized plant sources of protein and fat was associated with a
lower risk of mortality among general populations (13, 17, 20,
33), while the unhealthy LCD rich in animal sources of protein
and fat was associated with a higher risk of CVD events and
mortality (13, 20, 21). Therefore, it is necessary to take into
consideration the quality and sources of macronutrients when
investigating the associations of LCDs with health outcomes.
In patients with diabetes, glucose control is crucial to pre-
vent long-term vascular events and mortality (34, 35). Previous
clinical studies have found that adherence to an LCD could sig-
nicantly improve glycemic control, insulin sensitivity, HbA1c,
and dyslipidemia among patients with T2D in a relatively short
intervention period (9, 11, 36-39). Although an LCD has been
recommended by a US Consensus Report for glycemic control
among people with diabetes (22), the long-term safety and ef-
fect of LCD on the health outcomes of individuals with T2D
are still poorly understood. Whether different quality and com-
ponents of macronutrients in an LCD could yield diverse ef-
fects among diabetes is unclear. To ll these knowledge gaps,
our study for the rst time explored the associations of 3 dif-
ferent LCDs patterns with mortality among 5677adults with
T2D. We found that a healthy LCD, dened as lower intake of
low-quality carbohydrates, and higher intake of planted-based
protein and polyunsaturated fat, was signicantly associated
with lower level of CRP and higher level of HDL, and a lower
risk of mortality among individuals with T2D. Moreover,
sex, race/ethnicity, obesity status, and diabetes duration did
not modify the association of LCD score with mortality.
Unexpectedly, unhealthy LCD score was positively associated
with HDL level. Although previous studies have suggested
positive associations of very-low-carbohydrate ketogenic diets
or LCD scores with HDL levels (40, 41), more studies are war-
ranted to conrm this nding. In addition, given that lower
mortality was observed when energy from carbohydrates was
theoretically replaced by that from plant-based protein or
PUFAs, it is of great importance to verify the potential benet
of diet pattern in future intervention studies. Overall, our nd-
ings indicated that health benets of moderate LCD among
diabetic patients depend both on the quantity and quality of
macronutrients in thediet.
The potential mechanisms underlying the observed associ-
ations in our study might be explained by the following bio-
logical plausibility. Ahealthy LCD emphasizes higher intakes
of plant-based protein, and unsaturated fat and lower intakes
of animal-based foods, which have been shown to have favor-
able effects on cardiometabolic disease and traits (42, 43). In
addition, plant-based foods and their bioactive components
including ber, vitamins and minerals, and phytochemicals
may be involved in the associations between the healthy LCD
scores and mortality by creating a healthy gut-microbiota
environment (44, 45). Furthermore, greater intake of plant-
based foods such as fruits, vegetables, and nuts could reduce
the dietary acid load, which might improve proinammation
and endothelial dysfunction (46-48). Nevertheless, more
studies are warranted to clarify the underlying mechanisms.
To our best knowledge, the present study is among the rst
prospective studies to explore associations between different pat-
terns of moderate LCD and mortality among patients with dia-
betes. Moreover, the use of a nationally representatively sample
of US adults with diabetes could facilitate the generalization of
the ndings. In addition, because of the comprehensive data of
Figure 1. Multivariable adjusted HRs of total mortality by isocaloric replacing 2% of energy from total, unhealthy, or healthy carbohydrate with
specific macronutrients. HRs were adjusted for age (continuous), sex (male, or female), race/ethnicity (non-Hispanic White, non-Hispanic Black,
Mexican American, or other), body mass index (< 25, 25-30, or ≥ 30), education level (< high school, high school or equivalent, or ≥ college), family
income-poverty ratio (0-1.0, 1.0-3.0, or > 3.0), drinking status (nondrinker, low-to-moderate drinker, or heavy drinker), smoking status (never smoker,
former smoker, or current smoker), and leisure-time moderate-to-vigorous physical activity (0, 1-3, or > 3 times/week), duration of diabetes (< 3, 3-10,
or ≥ 10years), diabetes medication use (none, only oral medication, insulin, or others), glycated hemoglobin A1c (< 7.0% or ≥ 7.0%), total energy intake
(continuous), dietary cholesterol (in quartiles), self-reported hypertension (yes or no), hypercholesterolemia (yes or no), CVD (yes or no), and cancer (yes,
or no), and percentage of energy from remaining macronutrients where appropriate (animal or plant-based protein, healthy or unhealthy carbohydrates,
PUFAs, MUFAs, and SAFs, all continuous). Carbs, carbohydrates; HR, hazard ratio; MUFAs, monounsaturated fatty acids; PUFAs, polyunsaturated fatty
acids; SAFs, saturated fatty acids.
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022
e2708 The Journal of Clinical Endocrinology & Metabolism, 2022, Vol. 107, No. 7
NHANES, the present study was able to account for a multitude
of potential confounding factors. However, several limitations
should be mentioned as well. First, the self-reported diet data
using 2 24-hour recalls was subject to measurement error, al-
though we applied the National Cancer Institute method to es-
timate distribution of usual macronutrient intake (25). Second,
diet was assessed only at baseline and patients with diabetes
may have changed their diet during follow-up, despite it being
more likely to cause nondifferential misclassication bias and
attenuate the true associations. Third, diabetes type could not
be distinguished in NHANES, although T2D accounts for more
than 90% of patients with diabetes. Fourth, severity of diabetes
cannot be completely controlled because of a lack of detailed
information, although the results did not signicantly change
when adjusting for diabetes medication use, diabetes duration,
and HbA1c levels. Fifth, multiple comparisons could not be ex-
cluded, although the results were largely similar after using the
false discovery rate approach for multiple comparison correc-
tion. Finally, residual confounding cannot be entirely ruled out
because of the nature of the observational studydesign.
Conclusion
In a nationally representative sample of US adults, we found
that healthy LCD score was signicantly associated with a
lower risk of mortality among patients with T2D. Our nd-
ings indicate that adopting a well-balanced, moderate LCD
that emphasizes good quality and food sources of macronu-
trients may aid in preventing premature death among people
withT2D.
Acknowledgment
We would like to thank each author for your contributions.
FinancialSupport
G.L.was supported by the National Nature Science Foundation
of China (grant No. 82073554), a National Nutrition Science
Research Grant (No. CNS-NNSRG2021-10), the Hubei
Province Science Fund for Distinguished Young Scholars
(No. 2021CFA048), and the Fundamental Research Funds
for the Central Universities (No. 2021GCRC076). A.P. was
supported by the National Nature Science Foundation
of China (grant Nos. 81930124 and 82021005) and the
Fundamental Research Funds for the Central Universities
(No. 2021GCRC075). T.G. was supported by the China
Postdoctoral Science Foundation (No. 2021M691129).
Author Contributions
G.L.conceived the study design. Z.W. and Z.L.S.conducted
the analyses. Z.W.wrote the rst draft of the paper. All au-
thors contributed to the interpretation of the results and
critical revision of the manuscript for important intellectual
content. All authors approved the nal version of the manu-
script. G.L.is the guarantor of this work and, as such, had full
access to all the data in the study and takes responsibility for
the integrity of the data and the accuracy of the data analysis.
Disclosures
All authors have nothing to disclose.
Data Availability
Data described in the manuscript, code book, and analytic
code will be made available on request pending application
and approval from the corresponding author.
References
1. ZimmetP, AlbertiKG, MaglianoDJ, BennettPH. Diabetes mellitus
statistics on prevalence and mortality: facts and fallacies. Nat Rev
Endocrinol. 2016;12(10):616-622.
2. Saeedi P, Salpea P, Karuranga S, et al. Mortality attributable to
diabetes in 20-79 years old adults, 2019 estimates: results from
the International Diabetes Federation Diabetes Atlas, 9thedition.
Diabetes Res Clin Pract. 2020;162:108086.
3. ZhengY, LeySH, Hu FB. Global aetiology and epidemiology of
type 2 diabetes mellitus and its complications. Nat Rev Endocrinol.
2018;14(2):88-98.
4. ForouhiNG, MisraA, MohanV, TaylorR, YancyW. Dietary and
nutritional approaches for prevention and management of type 2
diabetes. BMJ. 2018;361:k2234.
5. Ludwig DS, WillettWC, Volek JS, Neuhouser ML. Dietary fat:
from foe to friend? Science. 2018;362(6416):764-770.
6. Hession M, RollandC, KulkarniU, WiseA, BroomJ. Systematic
review of randomized controlled trials of low-carbohydrate vs.
low-fat/low-calorie diets in the management of obesity and its
comorbidities. Obes Rev. 2009;10(1):36-50.
7. Clifton PM, CondoD, KeoghJB. Long term weight maintenance
after advice to consume low carbohydrate, higher protein diets—a
systematic review and meta analysis. Nutr Metab Cardiovasc Dis.
2014;24(3):224-235.
8. GoldenbergJZ, DayA, BrinkworthGD, etal. Efcacy and safety of
low and very low carbohydrate diets for type 2 diabetes remission:
systematic review and meta-analysis of published and unpublished
randomized trial data. BMJ. 2021;372:m4743.
9. Tay J, Luscombe-MarshND, ThompsonCH, etal. Comparison of
low- and high-carbohydrate diets for type 2 diabetes management:
a randomized trial. Am J Clin Nutr. 2015;102(4):780-790.
10. HallKD, Chung ST. Low-carbohydrate diets for the treatment of
obesity and type 2 diabetes. Curr Opin Clin Nutr Metab Care.
2018;21(4):308-312.
11. SantosFL, EstevesSS, daCostaPereiraA, YancyWS Jr, NunesJPL.
Systematic review and meta-analysis of clinical trials of the effects
of low carbohydrate diets on cardiovascular risk factors. Obes Rev.
2012;13(11):1048-1066.
12. Lagiou P, Sandin S, Weiderpass E, et al. Low carbohydrate-high
protein diet and mortality in a cohort of Swedish women. J Intern
Med. 2007;261(4):366-374.
13. Fung TT, vanDam RM, HankinsonSE, StampferM, WillettWC,
HuFB. Low-carbohydrate diets and all-cause and cause-specic mor-
tality: two cohort studies. Ann Intern Med. 2010;153(5):289-298.
14. Trichopoulou A, Psaltopoulou T, Orfanos P, Hsieh CC,
Trichopoulos D. Low-carbohydrate-high-protein diet and long-
term survival in a general population cohort. Eur J Clin Nutr.
2007;61(5):575-581.
15. Nakamura Y, Okuda N, Okamura T, et al; NIPPON DATA
Research Group. Low-carbohydrate diets and cardiovascular
and total mortality in Japanese: a 29-year follow-up of NIPPON
DATA80. Br J Nutr. 2014;112(6):916-924.
16. Nilsson LM, Winkvist A, Eliasson M, et al. Low-carbohydrate,
high-protein score and mortality in a northern Swedish population-
based cohort. Eur J Clin Nutr. 2012;66(6):694-700.
17. HaltonTL, WillettWC, Liu S, et al. Low-carbohydrate-diet score
and the risk of coronary heart disease in women. N Engl J Med.
2006;355(19):1991-2002.
18. ReynoldsA, MannJ, CummingsJ, WinterN, MeteE, TeMorengaL.
Carbohydrate quality and human health: a series of systematic
reviews and meta-analyses. Lancet. 2019;393(10170):434-445.
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022
The Journal of Clinical Endocrinology & Metabolism, 2022, Vol. 107, No. 7 e2709
19. BaoW, LiS, ChavarroJE, etal. Low carbohydrate-diet scores and
long-term risk of type2 diabetes among women with a history of
gestational diabetes mellitus: a prospective cohort study. Diabetes
Care. 2016;39(1):43-49.
20. ShanZ, GuoY, HuFB, LiuL, QiQ. Association of low-carbohydrate
and low-fat diets with mortality among US adults. JAMA Intern
Med. 2020;180(4):513-523.
21. LiS, FlintA, PaiJK, etal. Low carbohydrate diet from plant or an-
imal sources and mortality among myocardial infarction survivors.
J Am Heart Assoc. 2014;3(5):e001169.
22. Evert AB, Dennison M, Gardner CD, Garvey WT, Lau KHK,
MacLeod J, etal. Nutrition therapy for adults with diabetes or
prediabetes: a consensus report. Diabetes Care. 2019;42(5):731-754.
23. National center for health statistics. survey methods and ana-
lytic guidelines. Accessed April 23, 2022. https://wwwn.cdc.gov/
nchs/nhanes/AnalyticGuidelines.aspx#sample-design.
24. WanZ, ShanZ, GengT, etal. Supplementary data for “Associations of
moderate low-carbohydrate diets with mortality among patients with
type 2 diabetes: a prospective cohort study.” Deposited on February
18, 2022. Figshare. 2022. doi:10.6084/m9.gshare.19193672.
25. Tooze JA, Midthune D, Dodd KW, et al. A new statistical
method for estimating the usual intake of episodically consumed
foods with application to their distribution. J Am Diet Assoc.
2006;106(10):1575-1587.
26. MatthewsDR, HoskerJP, RudenskiAS, NaylorBA, TreacherDF,
Turner RC. Homeostasis model assessment: insulin resistance
and beta-cell function from fasting plasma glucose and insulin
concentrations in man. Diabetologia. 1985;28(7):412-419.
27. KulldorffM, Sinha R, ChowWH, RothmanN. Comparing odds
ratios for nested subsets of dietary components. Int J Epidemiol.
2000;29(6):1060-1064.
28. JiaoJ, Liu G, ShinHJ, et al. Dietary fats and mortality among
patients with type 2 diabetes: analysis in two population based co-
hort studies. BMJ (Clin Res Ed). 2019;366:l4009.
29. MalikVS, HuFB. Popular weight-loss diets: from evidence to prac-
tice. Nat Clin Pract Cardiovasc Med. 2007;4(1):34-41.
30. DysonP. A review of low and reduced carbohydrate diets and weight
loss in type 2 diabetes. J Hum Nutr Diet. 2008;21(6):530-538.
31. MazidiM, KatsikiN, MikhailidisDP, SattarN, BanachM. Lower
carbohydrate diets and all-cause and cause-specic mortality: a
population-based cohort study and pooling of prospective studies.
Eur Heart J. 2019;40(34):2870-2879.
32. DehghanM, Mente A, ZhangX, et al; Prospective Urban Rural
Epidemiology (PURE) Study Investigators. Associations of fats and
carbohydrate intake with cardiovascular disease and mortality in
18 countries from ve continents (PURE): a prospective cohort
study. Lancet 2017;390(10107):2050-2062.
33. SeidelmannSB, Claggett B, ChengS, et al. Dietary carbohydrate
intake and mortality: a prospective cohort study and meta-analysis.
Lancet Public Health. 2018;3(9):e419-e428.
34. HirakawaY, ArimaH, ZoungasS, etal. Impact of visit-to-visit gly-
cemic variability on the risks of macrovascular and microvascular
events and all-cause mortality in type 2 diabetes: the ADVANCE
trial. Diabetes Care. 2014;37(8):2359-2365.
35. Zoungas S, Arima H, GersteinHC, etal; Collaborators on Trials
of Lowering Glucose (CONTROL) Group. Effects of inten-
sive glucose control on microvascular outcomes in patients with
type 2 diabetes: a meta-analysis of individual participant data
from randomised controlled trials. Lancet Diabetes Endocrinol.
2017;5(6):431-437.
36. DavisNJ, TomutaN, SchechterC, etal. Comparative study of the
effects of a 1-year dietary intervention of a low-carbohydrate diet
versus a low-fat diet on weight and glycemic control in type 2 dia-
betes. Diabetes Care. 2009;32(7):1147-1152.
37. SkytteMJ, SamkaniA, PetersenAD, etal. A carbohydrate-reduced
high-protein diet improves HbA1c and liver fat content in weight
stable participants with type 2 diabetes: a randomised controlled
trial. Diabetologia. 2019;62(11):2066-2078.
38. van Zuuren EJ, Fedorowicz Z, Kuijpers T, Pijl H. Effects of
low-carbohydrate- compared with low-fat-diet interventions
on metabolic control in people with type 2 diabetes: a system-
atic review including GRADE assessments. Am J Clin Nutr.
2018;108(2):300-331.
39. Sainsbury E, Kizirian NV, Partridge SR, Gill T, Colagiuri S,
GibsonAA. Effect of dietary carbohydrate restriction on glycemic
control in adults with diabetes: a systematic review and meta-
analysis. Diabetes Res Clin Pract. 2018;139:239-252.
40. BuenoNB, deMeloISV, deOliveiraSL, daRochaAtaideT. Very-
low-carbohydrate ketogenic diet v. low-fat diet for long-term
weight loss: a meta-analysis of randomised controlled trials. Br J
Nutr. 2013;110(7):1178-1187.
41. NakamuraY, UeshimaH, Okuda N, etal. Relationship of three
different types of low-carbohydrate diet to cardiometabolic risk
factors in a Japanese population: the INTERMAP/INTERLIPID
Study. Eur J Nutr. 2016;55(4):1515-1524.
42. ChiavaroliL, ViguilioukE, NishiSK, et al. DASH dietary pattern
and cardiometabolic outcomes: an umbrella review of systematic
reviews and meta-analyses. Nutrients. 2019;11(2):338.
43. Chiavaroli L, NishiSK, KhanTA, et al. Portfolio dietary pat-
tern and cardiovascular disease: a systematic review and
meta-analysis of controlled trials. Prog Cardiovasc Dis. 2018;
61(1):43-53.
44. GentileCL, WeirTL. The gut microbiota at the intersection of diet
and human health. Science. 2018;362(6416):776-780.
45. Sonnenburg JL, Bäckhed F. Diet-microbiota interactions as
moderators of human metabolism. Nature. 2016;535(7610):
56-64.
46. DaneshzadE, KeshavarzSA, QorbaniM, LarijaniB, BellissimoN,
AzadbakhtL. Association of dietary acid load and plant-based diet
index with sleep, stress, anxiety and depression in diabetic women.
Br J Nutr. 2020;123(8):901-912.
47. MozaffariH, Namazi N, LarijaniB, Bellissimo N, AzadbakhtL.
Association of dietary acid load with cardiovascular risk factors
and the prevalence of metabolic syndrome in Iranian women: a
cross-sectional study. Nutrition. 2019;67-68:110570.
48. TostiV, BertozziB, FontanaL. Health benets of the Mediterranean
diet: metabolic and molecular mechanisms. J Gerontol ABiol Sci
Med Sci. 2018;73(3):318-326.
Downloaded from https://academic.oup.com/jcem/article/107/7/e2702/6569394 by Huazhong University of Science and Technology user on 25 July 2022