Dietary carbohydrates and metabolic outcomes: assessing the totality, consistency
and quality of epidemiologic observations and clinical
interventions
Simin Liu Department of Epidemiology
Brown University Division of Endocrinology, Department of Medicine
Rhode Island Hospital
Overview
♦ A diet that affects glucose homeostasis may be fundamental to vascular health outcomes • Glycemic response, glycemic index, and glycemic load
• Supports from evolutionary theory, epidemiologic observations, and experimental evidence
– Functional measures to study diet and disease in human populations
– Biological intermediaries for diabetes and cardiovascular disease
– Sex, body weight, and age as important modifiers
♦ Conclusions • Integrative framework for research
• Preventive recommendation
Diets (high GL/insulin demand) Genes
Obesity
Insulin resistance
Hyperinsulinemia Hyperglycemia
Dyslipidemia+TG
- HDL
Hypertension:Na+ retention;
SNS+Contractility+
Hemodynamic changes:inflammation,
Impaired fibinolysisand tthrombosis
Relative insulin deficiency
↑ Glycation of LDL ↑ Sorbitol ↓ NO/vasodilatory response
Coronary Heart Disease
Pathogenesis of DM/CHD Related to Insulin Resistance
Liu, 1998; Liu and Manson, 2001
↓Gluco-recognition ↓ Beta-cell mass; ↑Amyloid deposit
Traditional Epidemiology: Common Risk Factors for Three Major Metabolic Diseases
Risk factors Type 2 DM CHD Colon Ca.
Age ↑↑↑ ↑↑↑ ↑↑↑ Tobacco ↑ ↑↑ ↑↑ Physical inactivity ↑↑ ↑↑ ↑↑ Obesity ↑↑↑↑↑↑↑ ↑↑ ↑↑ Sex?? Excess energy intake ↑? ↑? ↑? Saturated fat ↑ ? ↑? ↑? Red meat ↑? ↑? ↑ Refined carbohydrates ↑↑? ↑↑? ↑↑? Dietary fibers ↓↓? ↓↓ ↓? Fruits & vegetables ↓↓? ↓↓ ↓? Whole grains ↓↓ ↓↓ ↓? Nuts/legumes ↓↓? ↓↓ ↓? Moderate Alcohol ↓↓ ↓↓ ↑? ……
Modified from Giovannucci 1995 and Liu 1998
Sex differences related to vascular outcomes in prospective studies
Diabetes and risk of CVD mortality stronger in women
BMI and risk of type 2 diabetes stronger in women
Willett NEJM 1999; Huxley BMJ 2005
Dose-response relation between blood glucose and CHD risk (Levitan et al. Arch Int Med, 2004)
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
50 75 100 125 150 175 200
Plasma glucose (mg/dL)
Log
rela
tive
risk
Fasting
Postchallenge
Fasting
Postchallenge
The effect of decreasing PPG with acarbose, an -glucosidase inhibitor, on DM development in a multicenter double-blind, placebo-controlled, randomized trial Chiasson et al. Lancet 2002
The effect of decreasing PPG with acarbose, an -glucosidase inhibitor, on CVD development in a multicenter double-blind, placebo-controlled, randomized trial Chiasson et al. JAMA 2004
0!500!1000!1500!
Kilo
calo
ries
Con
sum
ed!
1! 2! 3! 4! 5!Time (hr)
Cumulative Voluntary Food Intake
High GI!Med GI!Low GI!
GI & Voluntary Food Intake Ludwig et al. Pediatrics 1999, 103:e261
Low GI diets improve triglycerides Low GI diet compared with high GI diet
-25 -20 -15 -10 -5 0 5
Frost et al 1994
Fontvieille et al 1992
Wolever et al 1992
Fontvieille et al 1988
Jenkins et al 1985
Jenkins et al 1987
Jarvi et al 1999
Brand et al 1991
25 type 2 , 12 wks
18 type 1 & 2, 5 wks
6 type 2, 6 wks
8 type 1, 3 wks
12 CHD, 4 wks
30 CHD, 4 wks
20 type 2, 24 days
Average change in 8 studies = -12% !*Include extensive data from Prof.Y.X. Yang Chinese CDC
Changes in HbA1c or fructosamine Low GI diets compared with high GI diets
-35 -30 -25 -20 -15 -10 -5 0
Frost et al 1994
Fontvieille et al 1992
Brand et al 1991
Wolever et al 1992
Wolever et al 1992
Jenkins et al 1988
Fontvielle 1988
Collier et al 1988
Gilbertson et al 2000
25 type 2 , 12 wks!18 type 1 & 2, 5 wks!
16 type 2, 12 wks!6 type 2, 6 wks!
15 type 2, 2 wks!
8 type 2, 4 weeks!
8 type 1, 3 weeks!
7 type 1 children, 6 wks!104 type 1 children, 12 mo!
!
Average difference in HbA1c in 9 studies = -11% !Adapted from Brand-Miller et al. 2001 !
Change in Glycemic Control Low vs high GI diets
• Glycemic control - 10% improvement for a 10 unit decrease in GI
• Comparison with other interventions
DCCT - 20% UKPDS - 11.5%
Insulin analogues - 2.5%
Acarbose - 6.5%
Adapted from Stephen Colagiuri Aust Nutr Soc 2001
80
90
100
110
120
130
140
150
160
1 2 3 4 5Quintile of Intake
Glycemic Index (p=0.03)
Carbohydrate (p=0.005)
Glycemic Load (p<0.001)
Fasting plasma TG concentrations by GL, GI and carbohydrate intake
Liu et al. AJCN 2001
Nurses’ Health Study
Fast
ing
Trig
lyce
rides
(mg/
dl)
12.039b
Fasting Plasma TG Levels by Dietary Glycemic Load
Postmenopausal Women with Different BMI’s
0
50
1 00
1 50
2 00
Q u in t il e o f E n ergy-ad ju s te d G lyce mic L oad
Q1 Q2 Q3 Q4 Q5
Fasting TG (mg/dl)
Liu et al. AJCN, 2001
BMI > 25BMI < 25
Adjusted geometric mean plasma concentrations of high-sensitivity C-reactive protein (hs-CRP) by quintiles (Q1–Q5) of energy-adjusted dietary glycemic load
in 244 women in 2 BMI categories
Liu S et al. Am J Clin Nutr 2002;75:492-498
Subsequently confirmed by Rhodes et al in a randomized trial of overweight pregnant women AJCN 2010
30
35
40
45
50
55
60
65
1 2 3 4 5
Quintile of dietary glycemic index
High
den
sity
lipo
prot
ein
chol
este
rol
leve
l (m
g/dL
)
Normal weightOver weight
30
35
40
45
50
55
60
65
1 2 3 4 5
Quintile of dietary glycemic load
High
den
sity
lipo
prot
ein
chol
este
rol
leve
l (m
g/dL
)
40
60
80
100
120
140
160
180
1 2 3 4 5
Quintile of dietary glycemic index
Trig
lyce
ride
leve
l (m
g/dL
)
40
60
80
100
120
140
160
180
1 2 3 4 5
Quintile of dietary glycemic loadTr
igly
cerid
ele
vel (
mg/
dL)
P for interaction = 0.87 P for interaction = 0.0001
P for interaction = 0.81 P for interaction = 0.32
30
35
40
45
50
55
60
65
1 2 3 4 5
Quintile of dietary glycemic index
High
den
sity
lipo
prot
ein
chol
este
rol
leve
l (m
g/dL
)
Normal weightOver weight
30
35
40
45
50
55
60
65
1 2 3 4 5
Quintile of dietary glycemic load
High
den
sity
lipo
prot
ein
chol
este
rol
leve
l (m
g/dL
)
40
60
80
100
120
140
160
180
1 2 3 4 5
Quintile of dietary glycemic index
Trig
lyce
ride
leve
l (m
g/dL
)
40
60
80
100
120
140
160
180
1 2 3 4 5
Quintile of dietary glycemic loadTr
igly
cerid
ele
vel (
mg/
dL)
P for interaction = 0.87 P for interaction = 0.0001
P for interaction = 0.81 P for interaction = 0.32
Geometric mean high density lipoprotein cholesterol and triglyceride levels*
*Adjusted for age, smoking (current, past, never), exercise (4 category), hormone replacement therapy use (current, past, never), family history of MI, history of diabetes, body mass index (4 category), and intakes of total energy, alcohol intake (4 category), and quintile of total fat, protein, cholesterol, folate, and magnesium intake. Dietary glycemic load additionally adjusted for quintile of fiber intake
Relation between HDL-cholesterol concentration and glycemic index in men and women (Frost et al, 1999)
Mean concentrations of HDL by glycemic index among men and women aged 20+ years, NHANES III,
1988-1994 (Ford and Liu, Arch Int Med 2000)
11.11.21.31.41.51.6
<76%76-79%80-83%84-87%
>87%
<76%76-79%80-83%84-87%
>87%
Glycemic index quintiles
HDL cholesterol
(mmol/l)
Adjusted for age, race or ethnicity, education, smoking status, body mass index, alcohol intake, physical activity, percent kilocalories from protein (quintiles), percent kilocalories from fat (quintiles), total energy intake (quintiles).
Men
Women
P<0.0001
-200"
0"
200"
400"
600"
800"
1000"
1200"BX
D-20/TyJ"
AXB-19/PgnJ"
BXHA1"
BXD-34/TyJ"
C57BL/6J"
BXA-16/PgnJ"
BXD49"
BXA-24/PgnJ"
BXD-40/TyJ"
NON/LtJ"
BXA-14/PgnJ"
BXD-13/TyJ"
BXD79"
CXB
-6/ByJ"
BXD-24/TyJ"
CXB
H"
BXA-11/PgnJ"
BXD51"
C3H
/HeJ"
AXB-10/PgnJ"
BXD66"
BXH-6/TyJ"
MA/MyJ"
BUB/BnJ"
CBA
/J"
BXH-19/TyJ"
BXD56"
CXB
-12/HiAJ"
BXD-6/TyJ"
BXD-38/TyJ"
AXB-15/PgnJ"
RIIIS/J"
AXB-18/PgnJ"
BXA-2/PgnJ"
DBA
/2J"
BXHB2"
BXH-9/TyJ"
BXD60"
BXD-31/TyJ"
BXD-12/TyJ"
BXD-14/TyJ"
FVB/NJ"
BXD-19/TyJ"
BTBR
T<+>tf/J"
AXB-6/PgnJ"
A/J"
BXA-7/PgnJ"
C57BLKS
/J"
BXD43"
BXD-32/TyJ"
BXD55"
BALB/cJ"
BXD-1/TyJ"
BXD-21/TyJ"
CXB
-11/HiAJ"
BXA-4/PgnJ"
BXA-1/PgnJ"
SEA/GnJ"
BXD-9/TyJ"
CXB
-13/HiAJ"
PL/J"
CXB
-7/ByJ"
SWR/J"
CXB
-3/ByJ"
% Total Adipo
se Growth
(0 to
8 W
eeks)
Male Female
Results: Diet-‐by-‐sex interac3on
Slide courtesy of Drs Tom Drake & Jake Lusis
Q1(Low)Q2Q3Q4Q5(High)0
0.25
0.5µmol/l
Mean concentrations of Carotenoids by glycemic index among 15270 US adults, NHANES III
Quintile of dietary glycemic index
lycopene
Lutein/zeax
beta-carotene
Crptoxanthin
α-carotene
Liu and Ford, 2002
Q1(Low)Q2Q3Q4Q5(High)20
25
30
35
40
45
50
Quintile Of Dietary Glycemic Index
Mean concentrations of serum Vitamin E and Vitamin C by glycemic index among 15270 US adults, NHANES III
Vitamin C (nmol/L)
Vitamin E (µmol/L)
Liu and Ford, 2002
Potential biochemical mechanisms High GI/GL feeding causes: ♦ Postprandial hyperglycemia & hyperinsulinemia -
(immediate responses) ♦ Counterregulatory hormonal responses which
stimulate appetite, FFA production, and possibly protein breakdown – (late responses)
♦ Shifts in substrate utilization away from fat towards carbohydrates
♦ Increased enzymatic capacities for carbohydrate oxidation and lipogenesis, and decreased enzymatic capacity of fat oxidation
Adjusted estimates of relative risk of type 2 diabetes according to GL by sex from a meta-analysis of all prospective cohorts with 7.5 million years of
followup (up to August 2012)
Livesay et al. Am J Clin Nutr 2013
Multivariate relative risk of CHD by body mass index and glycemic load
1.11
2.03 1.97
0.941.2
1.74
1.00 1.05 1.42
0.000.501.001.502.002.50
RR
< 23 23-29 >29Tertile 1
Tertile 2Tertile 3
BMI
GL
Test for interaction, P < 0.01
Liu et al 2000
Different constituents of a low GI/GL diet
♦ Fruits and vegetables
♦ Whole grains
♦ Nuts
♦ Legumes (peas, beans, soybeans)
♦ Fish
♦ Dairy foods
Conclusion ♦ Due to our genetic mal-adaptation to westernized lifestyle/
environment, we are increasingly becoming a metabolically efficient population
♦ Substantial evidence indicates significant biological importance and clinical utilities of the GI/GL concepts, in that • 1) carbohydrate-containing foods differ in their abilities to raise plasma
glucose and insulin (Level A), • 2) diets characterized by high GL adversely affect metabolic intermediates
(Level B), • 3) may increase risk of vascular outcomes (diabetes and cardiovascular
diseases) especially among those who are prone to insulin resistance (Level B and C)
♦ Can we conclusively demonstrate the efficacy of glucose
homeostasis diet on vascular health in randomized trials: • Not feasible • Not ethical • Not necessary
Conclusion ♦ When thinking about diet, which is only one aspect of our
westernized environment, we need to keep in mind what kind of people we have become metabolically.
• Consider sex, body weight, and age as important modifiers when
assessing diet-disease relation and making dietary recommendation – Reduce heterogeneity and improve biological understanding of the “glucose homeostasis” diet
• Regarding the clinical utility of the GI concept, – It is better than the “simple/complex” in classifying carbohydrates – Does go beyond individual food-based guidelines, even in the context
of caloric density and nutrient composition – GI is most useful for ranking high/dense-carbohydrate foods