Content uploaded by Fumiaki Imamura
Author content
All content in this area was uploaded by Fumiaki Imamura on Aug 04, 2016
Content may be subject to copyright.
Content uploaded by Jason H Y Wu
Author content
All content in this area was uploaded by Jason H Y Wu on Jul 19, 2016
Content may be subject to copyright.
RESEARCH ARTICLE
Effects of Saturated Fat, Polyunsaturated
Fat, Monounsaturated Fat, and Carbohydrate
on Glucose-Insulin Homeostasis: A
Systematic Review and Meta-analysis of
Randomised Controlled Feeding Trials
Fumiaki Imamura
1
*, Renata Micha
2
, Jason H. Y. Wu
3
, Marcia C. de Oliveira Otto
4
, Fadar
O. Otite
5
, Ajibola I. Abioye
6
, Dariush Mozaffarian
2
1Medical Research Council Epidemiology Unit, Institute of Metabolic Science, University of Cambridge
School of Clinical Medicine, Cambridge Biomedical Campus, Cambridge, United Kingdom, 2Tufts Friedman
School of Nutrition Science & Policy, Boston, Massachusetts, United States of America, 3George Institute
for Global Health, The University of Sydney, Sydney Medical School, Camperdown, Australia, 4Department
of Epidemiology, Human Genetics & Environmental Sciences, The University of Texas Health Science
Center at Houston, Houston, Texas, United States of America, 5Department of Neurology, University of
Miami Miller School of Medicine/Jackson Memorial Hospital, Miami, Florida, United States of America,
6Department of Global Health and Population, Harvard T. H. Chan School of Public Health, Boston,
Massachusetts, United States of America
*fumiaki.imamura@mrc-epid.cam.ac.uk
Abstract
Background
Effects of major dietary macronutrients on glucose-insulin homeostasis remain controver-
sial and may vary by the clinical measures examined. We aimed to assess how saturated
fat (SFA), monounsaturated fat (MUFA), polyunsaturated fat (PUFA), and carbohydrate
affect key metrics of glucose-insulin homeostasis.
Methods and Findings
We systematically searched multiple databases (PubMed, EMBASE, OVID, BIOSIS, Web-
of-Knowledge, CAB, CINAHL, Cochrane Library, SIGLE, Faculty1000) for randomised con-
trolled feeding trials published by 26 Nov 2015 that tested effects of macronutrient intake on
blood glucose, insulin, HbA1c, insulin sensitivity, and insulin secretion in adults aged 18
years. We excluded trials with non-isocaloric comparisons and trials providing dietary
advice or supplements rather than meals. Studies were reviewed and data extracted inde-
pendently in duplicate. Among 6,124 abstracts, 102 trials, including 239 diet arms and
4,220 adults, met eligibility requirements. Using multiple-treatment meta-regression, we
estimated dose-response effects of isocaloric replacements between SFA, MUFA, PUFA,
and carbohydrate, adjusted for protein, trans fat, and dietary fibre. Replacing 5% energy
from carbohydrate with SFA had no significant effect on fasting glucose (+0.02 mmol/L,
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 1/18
a11111
OPEN ACCESS
Citation: Imamura F, Micha R, Wu JHY, de Oliveira
Otto MC, Otite FO, Abioye AI, et al. (2016) Effects of
Saturated Fat, Polyunsaturated Fat,
Monounsaturated Fat, and Carbohydrate on Glucose-
Insulin Homeostasis: A Systematic Review and Meta-
analysis of Randomised Controlled Feeding Trials.
PLoS Med 13(7): e1002087. doi:10.1371/journal.
pmed.1002087
Academic Editor: Ronald C. W. Ma, Chinese
University of Hong Kong, CHINA
Received: August 3, 2015
Accepted: June 10, 2016
Published: July 19, 2016
Copyright: © 2016 Imamura et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any
medium, provided the original author and source are
credited.
Data Availability Statement: Study materials are
deposited to the repository of the University of
Cambridge MRC Epidemiology Unit, the institutional
repository (https://www.repository.cam.ac.uk/handle/
1810/245351). The materials are identifiable by
searching the authors' names or the title.
Funding: FI received support from the Medical
Research Council Epidemiology Unit Core Support
(MC_UU_12015/5). DM received funding from The
National Institute of Health in the United States (R01
95% CI = -0.01, +0.04; ntrials = 99), but lowered fasting insulin (-1.1 pmol/L; -1.7, -0.5; n=
90). Replacing carbohydrate with MUFA lowered HbA1c (-0.09%; -0.12, -0.05; n=23),2h
post-challenge insulin (-20.3 pmol/L; -32.2, -8.4; n= 11), and homeostasis model assessment
for insulin resistance (HOMA-IR) (-2.4%; -4.6, -0.3; n= 30). Replacing carbohydrate with
PUFA significantly lowered HbA1c (-0.11%; -0.17, -0.05) and fasting insulin (-1.6 pmol/L; -2.8,
-0.4). Replacing SFA with PUFA significantly lowered glucose, HbA1c, C-peptide, and
HOMA. Based on gold-standard acute insulin response in ten trials, PUFA significantly
improved insulin secretion capacity (+0.5 pmol/L/min; 0.2, 0.8) whether replacing carbohy-
drate, SFA, or evenMUFA. No significant effects of any macronutrient replacements were
observed for 2 h post-challenge glucose or insulin sensitivity (minimal-model index). Limita-
tions included a small number of trials for some outcomes and potential issues of blinding,
compliance, generalisability, heterogeneity due to unmeasured factors, and publication bias.
Conclusions
This meta-analysis of randomised controlled feeding trials provides evidence that dietary
macronutrients have diverse effects on glucose-insulin homeostasis. In comparison to car-
bohydrate, SFA, or MUFA, most consistent favourable effects were seen with PUFA, which
was linked to improved glycaemia, insulin resistance, and insulin secretion capacity.
Author Summary
Why Was This Study Done?
•Effects of dietary fat and carbohydrate on metabolic health have been controversial,
leading to confusion about specific dietary guidelines and priorities.
•To date there has not been a systematic evaluation of all available evidence to quantify
the effects of dietary fat (saturated, monounsaturated, and polyunsaturated fat), and car-
bohydrate on various markers mediating the development of diabetes, including blood
sugar, insulin sensitivity, and ability to produce insulin.
What Did the Researchers Do and Find?
•We systematically identified and summarized findings of 102 randomised controlled tri-
als, including a total of 4,660 participants, that provided meals varying in the types and
levels of fat and carbohydrate to study participants and evaluated how such variations
affected various measures of blood glucose control, insulin sensitivity, and ability to pro-
duce insulin.
•The findings suggest exchanging dietary carbohydrate with saturated fat does not appre-
ciably influence markers of blood glucose control.
•On the other hand, substituting carbohydrate and saturated fat with a diet rich in unsat-
urated fat, particularly polyunsaturated fat, was beneficial for the regulation of blood
sugar.
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 2/18
HL085710). The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing Interests: I have read the journal's policy
and the authors of this manuscript have the following
competing interests: DM reports ad hoc honoraria or
consulting from Boston Heart Diagnostics, Haas
Avocado Board, Astra Zeneca, GOED, DSM, and Life
Sciences Research Organization; chapter royalties
from UpToDate; and scientific advisory board,
Elysium Health. Harvard University has been
assigned patent US8889739 B2, listing DM as one of
three co-inventors, for use of trans-palmitoleic acid in
identifying and treating metabolic disease. RM and
JHYW received research support from Unilever R&D
(project reference number MA-2015-01161) for work
on fatty acid biomarkers and incident cardiometabolic
diseases.
Abbreviations: AIR, acute insulin response; BMI,
body-mass index; CHO, carbohydrate; HOMA-IR,
homeostasis model assessment for insulin
resistance; ISI, insulin-sensitivity index; MUFA,
monounsaturated fat; PUFA, polyunsaturated fat;
SFA, saturated fat.
What Do These Findings Mean?
•These findings may help inform scientists, clinicians, and the public on dietary priorities
related to dietary fats and carbohydrates and metabolic health.
•This investigation suggests that consuming more unsaturated fats in place of either car-
bohydrates or saturated fats will help improve blood glucose control. Sole emphasis on
lowering consumption of carbohydrates or saturated fats would not be optimal.
•Translated to foods, these findings support benefits of increasing consumption of vege-
table oils and spreads, nuts, fish, and vegetables rich in unsaturated fats (e.g., avocado),
in place of either animal fats or refined grains, starches, and sugars.
Introduction
The prevalence of insulin resistance and type 2 diabetes is rising sharply in nearly all nations
globally [1,2], highlighting the need for broad preventive therapies. Diet is a cornerstone of pre-
vention and treatment in all major guidelines [3,4]. Dietary guidelines on macronutrient intake
to improve glucose-insulin profiles and reduce or prevent type 2 diabetes generally recommend
increasing foods rich in monounsaturated fat (MUFA) and reducing saturated fat (SFA) [3–6].
Yet these guidelines have also emphasized the major gaps in established evidence for effects of
dietary fats and carbohydrate on glucose-insulin homeostasis, including uncertainty as to
whether benefits of MUFA in some trials were confounded by caloric restriction and limited
evidence on effects of either polyunsaturated fat (PUFA) or SFA [3–7]. Understanding the role
of dietary macronutrients in glucose-insulin control is crucial to establishing informed guide-
lines for clinical providers and policy-makers around the world.
Prior knowledge has been limited by several factors, including focus on limited metrics to
assess glucose-insulin homeostasis (e.g., fasting glucose alone), rather than studying multiple
relevant outcomes, such as HbA1c, fasting insulin, insulin resistance, insulin secretion capacity,
and post-challenge measures [8]; insufficient statistical power in many smaller trials to confirm
important effects; and difficulties in evaluating results of individual trials due to multiple and
varying changes in several macronutrients simultaneously [8–11]. Due to these challenges, the
effects of dietary fats and carbohydrate on glucose-insulin homeostasis remains uncertain [8].
To address these critical gaps in knowledge, we performed a systematic review and dose-
response meta-regression of randomised controlled feeding trials that tested the effects of iso-
caloric diets with differing composition of dietary macronutrients on multiple key metrics of
fasting and post-challenge glucose-insulin homeostasis that represent degrees of glycaemia,
insulin resistance, and insulin secretion capacity.
Methods
Eligibility Criteria and Literature Search
We developed the protocol (S1 Text) and conducted this study following Preferred Reporting
Items for Systematic reviews and Meta-Analysis (PRISMA) guidelines [12](S2 Text). Details
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 3/18
of literature search and data preparation are provided in S3 Text. We systematically searched
for randomised controlled feeding trials in adults (aged 18 y) examining diets varying in
composition of specific fats and/or carbohydrate. Eligibility criteria included: provision of
meals; comparison of isocaloric interventions; and assessment of relevant glucose-insulin met-
rics. We focused on outcomes commonly assessed in clinical research or practice [8,13], includ-
ing fasting glucose, fasting insulin, haemoglobin A1c (HbA1c), homeostasis model assessment
for insulin resistance (HOMA-IR, a fasting or post-challenge measure of insulin resistance cal-
culated from glucose and insulin), C-peptide, 2 h post-oral-challenge glucose and insulin, and
intravenous-infusion measures of Minmod-based insulin-sensitivity index (ISI) and acute
insulin response (AIR) (gold-standard measures of insulin sensitivity and β-cell function,
respectively) [8,13]. Study exclusions were insufficient information on macronutrient composi-
tion or glycaemic outcomes, studies of supplements or dietary advice only, and studies of acute
(single meal) post-prandial effects only. We searched PubMed, EMBASE, OVID, BIOSIS,
Web-of-Knowledge, CAB, CINAHL, Cochrane Library, SIGLE, and Faculty 1000, without lan-
guage restriction, for all publications up until 26 November 2015. Search terms included each
of the dietary macronutrients and metabolic measurements of interest. Titles and abstracts
were screened by one investigator for eligibility; the full-text of potentially eligible reports was
reviewed independently and in duplicate. Citation lists of included articles and identified prior
reviews were similarly searched for relevant articles.
Data Extraction
For each included trial, information was extracted independently (by FI, RM, JHYW, MCdOO,
FOO, AIA) and in duplicate on first author, publication year, location, design, participant char-
acteristics, dietary intervention, outcomes, compliance, and loss to follow-up. Any required
information that was not reported was obtained by direct contact with authors (27 of 66
responded), other publications from the same trial, or trial-registry websites when available.
Certain values were estimated using reported data: e.g., a mid-point was used if only a range
was presented for age or body-mass index (BMI); in one trial, the reported consumption of
rapeseed oil was combined with its macronutrient composition to estimate the intakes of spe-
cific dietary fats (S3 Text). Study quality was examined by using Jadad scale [14]: two authors
independently scored each of the 11 quality-related items, calculated total scores of the 11 com-
ponents and averaged two summed scores for each trial. Outcome measures presented in fig-
ures (e.g., insulin levels after glucose insulin) were digitalised to numeric information by two
authors (FI and MCdOO) using software (Dagra, Blue Leaf Software Ltd., Hamilton, New Zea-
land), and two values for a single estimate were averaged.
Meta-analysis
We evaluated the post-intervention values (means, standard errors) of trial arms as the primary
outcomes. Changes in outcome values from baseline to endpoint were not used because certain
procedures (intravenous tests) were often implemented only at endpoints and because baseline
values were more subject to bias due to a carry-over effect in a crossover trial. When values
were log-transformed, they were standardised to non-transformed values [15], except for
HOMA-IR, which was standardised to log-transformed values. Between-arm correlations in
trials using either crossover or Latin-square design were estimated and incorporated in meta-
analysis by using reported p-values and outcome measures based on the function of within-
individual correlations, interventional effects, their standard errors or deviations, and p-value
[15,16]. Missing information on covariates (trans fat, dietary fibre), within-trial correlations, or
precise post-intervention statistics (e.g., results expressed only as “p>0.05”; standard
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 4/18
deviations of post-intervention values [17]) was imputed with a multiple imputation approach
to incorporate the uncertainty in our estimation by generating ten imputed datasets and pool-
ing the estimates [18].
We estimated dose-response effects of replacement among carbohydrate, SFA, MUFA, and
PUFA using multiple-treatments meta-regression (command: SAS PROC GLIMMIX, SAS
Inc., North Carolina, United States) [19]. This meta-regression is an extension of a standard
inverse-variance weighted model, expressed as Y
ij
=I
i
+ SFA
ij
×β
SFA
+ MUFA
ij
×β
MUFA
+ PUFA
ij
×β
PUFA
+ Covariates
ijk
×β
k
+ε
ij,
modelling different macronutrients as multiple-
treatment variables (SFA
ij
, MUFA
ij
, and PUFA
ij
) of trial i’s arm j, as well as study-specific
intercepts (I
i
), arm-specific covariates k(protein, trans fat, dietary fibre), arm-specific standard
errors of post-intervention values (ε
ij
, standard deviation
ij
/pn
ij
), and their within-trial corre-
lations based on trial design (r = 0.01–0.99 in crossover or Latin-square trials; r = 0 in parallel
trials) specified in variance-covariance structure of ε
ij,
[16,20]. We used fixed-effects models,
assessing both main effects and sources of heterogeneity (see below) [21]. In a stratum with a
small number of trials, the model with five fixed-effects parameters was not fitted. We recog-
nized the divergence of opinion on optimal weighting methods in the presence of statistical
heterogeneity; in post hoc sensitivity analysis, we carried out random-effects meta-analyses
(three τ
2
for β
SFA
,β
MUFA
, and β
PUFA
, assumed to be independent) following stratification or
restriction by significant sources of heterogeneity.
We evaluated SFA, MUFA, and PUFA (% energy) as main treatments, in comparison to iso-
caloric replacement with carbohydrate, by including each of these dietary fats in the model as
well as intakes of protein (% energy) and trans fat (% energy) [9–11]. Effects of interchanging
different fats were estimated by subtraction of corresponding regression coefficients (i.e.,
β
MUFA
–β
SFA
,β
PUFA
–β
SFA
,β
PUFA
–β
MUFA
)[20]. Because trans fat is a potential confounder not
included in other meta-analyses of dietary fats [9,10], we extracted information on trans fat
consumption in all trials reporting such data and imputed it within the remaining trials, with
sensitivity analyses examining the effects of different methods for imputation and adjustment
(S3 Text). To account for differences in carbohydrate quality between arms and trials, we also
adjusted for dietary fibre intake (g/1,000 kcal) in each arm.
Assessment of Heterogeneity, Sensitivity Analyses, and Small Study
Bias
Hypothesizing that differences in effects of dietary macronutrients on fasting glucose, fasting
insulin, HbA1c, and HOMA-IR would not be at random, we explored pre-specified potential
sources of heterogeneity. These included study mean age (years), sex (% men), location (US/
Canada, Europe/Australia, Asia), design (parallel, crossover/Latin-square), intervention dura-
tion (weeks), diabetes (yes/no), caloric restriction (yes/no), drop-out rate (%), participant
blinding of meals provided (yes/no), mean BMI (kg/m
2
), mean baseline fasting glucose (mmol/
L), mean fibre intake (g/1,000 kcal), mean weight change during intervention (kg), and study
quality score (points). In post hoc analyses, we explored heterogeneity by extent of provision of
all daily meals (full/partial). Each characteristic was tested as a potential source of heterogeneity
by testing a standard Q-statistics for stratum-specific effects on the selected outcome for
exchanging carbohydrate with SFA, MUFA, or PUFA, exchanging SFA with MUFA or PUFA,
and exchanging MUFA with PUFA. For stratification by continuous variables, the median
value across studies was used. To avoid false positive findings due to multiple testing of these
exploratory interactions on the four outcomes, the α= 0.05 was adjusted for the family-wise
false-discovery rate [22]. To minimize additional multiple comparisons, we explored potential
interactions for the other outcomes (2 h glucose, 2 h insulin, ISI, AIR) only for those
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 5/18
characteristics identified as significant sources of heterogeneity for fasting glucose, insulin,
HbA1c, or HOMA, again adjusted for the false-discovery rate. Due to limited power, we did
not explore heterogeneity for outcomes having ten or fewer trials (C-peptide).
We performed several sensitivity analyses for the main findings on fasting glucose, HbA1c,
and fasting insulin, including varying the estimated between-arm correlation in crossover trials
(S3 Text), repeating meta-analysis with and without adjustment for protein, fibre, and trans
fat; using different methods for imputing and adjusting for trans fat; and adjusting for total
caloric intake and for within-trial weight change to examine the potential mediating effect of
macronutrient composition on energy metabolism [23,24] and between-arm imbalance in
compliance to isocaloric intervention. In post hoc sensitivity analysis, we restricted to trials
with follow-up 4 wk (the median of all trials), which may be especially relevant for longer-
term measures such as HbA1c [25]; to trials using caloric-restriction, to explore whether this
altered overall findings; and to trials with primary aims of varying either SFA, MUFA, or
PUFA, to explore potential influence of combining trials with different original aims [9,20].
To assess publication bias or bias specific to small studies in multiple-treatment meta-
regression, we utilized influence analyses [15]. Meta-regressions were repeated after excluding
each single trial individually, with each new meta-regression finding plotted against the square
root of the excluded trial’s effective sample size, accounting for within-trial correlations [26].
The resulting plots were inspected visually for patterns of bias by trial size; using linear regres-
sion to determine whether observed deviations were statistically significant, analogous to
Egger’s test [15]; and using a non-parametric Wilcoxon rank test to examine whether estimates
were symmetrical around the main estimate.
Results
Of 6,124 identified abstracts, 102 trials met inclusion criteria, evaluating a total of 4,220 unique
subjects (45% male) across 239 dietary arms (Fig 1,Table 1,S1 Table, and S2 Table). Eleven tri-
als implemented oral glucose or meal tolerance tests to assess 2 h post-challenge glucose or
insulin; 13 trials, intravenous infusion tests to assess insulin sensitivity; and 10 trials, intrave-
nous tests to assess insulin secretion capacity. No trials reported significant energy imbalance
between arms after interventions. The average study quality was moderate to high (out of a
possible score range of 0 to 11, median: 8.0, range: 4 to 10.5; see S2 Table).
Fasting Glucose, HbA1c, and 2 h Glucose
Ninety-nine trials including 237 dietary arms evaluated fasting glucose. In pooled analysis,
each 5% energy exchange of carbohydrate with SFA, MUFA, or PUFA did not significantly
alter fasting glucose levels (p>0.16 each) (Table 2). Exchanges between SFA, MUFA, and
PUFA also did not alter fasting glucose (p>0.15 each), except for the replacement of SFA with
PUFA, which was linked to a decrease in fasting glucose levels (-0.04 mmol/L; 95% CI: -0.07,
-0.01; p= 0.028).
Among 23 trials including 54 dietary arms and assessing HbA1c, isocaloric replacement of
5% dietary energy from either carbohydrate or SFA with 5% dietary energy from either MUFA
or PUFA lowered HbA1c (p<0.001 each) (Table 2). In eleven trials assessing 2 h post-chal-
lenge glucose no significant effects of macronutrient exchanges were identified.
Insulin, Insulin Sensitivity, and Insulin Secretion
Ninety trials including 216 arms evaluated fasting insulin (Table 2). Compared with 5% dietary
energy from carbohydrate, 5% dietary energy from either SFA or PUFA reduced fasting insulin
by 1.1 pmol/L (0.6, 1.6; p= 0.001) and 1.6 pmol (0.4, 2.8; p= 0.015), respectively, while
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 6/18
replacement with MUFA had no significant effect (0.1 pmol/L; -0.03, 0.04; p= 0.001). How-
ever, replacement of carbohydrates with MUFA was linked to increased fasting insulin (+1.2
pmol/L; 0.6, 1.8; p= 0.001). In 11 trials evaluating 2 h post-challenge insulin, replacement of
carbohydrate or SFA with MUFA or PUFA did not significantly reduce the fasting insulin lev-
els; while replacing MUFA with carbohydrate significantly lowered 2 h insulin (-20.3 pmol/L;
-32.2, -8.4; p= 0.001). In 7 trials, consuming SFA in place of carbohydrate significantly
increased C-peptide (0.03 nmol/L; 0.00, 0.05; p= 0.024).
The effects on HOMA-IR of consuming MUFA or PUFA in place of carbohydrate or SFA
(30 trials) were generally similar to findings for fasting glucose, HbA1c, and 2 h insulin. For
example, consuming 5% energy from PUFA in place of carbohydrate or SFA lowered
HOMA-IR by 3.4% (0.8, 5.9%; p= 0.010) and 4.1% (1.6, 6.4%; p= 0.001), respectively.
Intravenous gold-standard measures of insulin sensitivity (ISI) and insulin secretion capac-
ity (AIR) were assessed in 13 trials and 10 trials, respectively (Table 2). No significant effects of
macronutrient replacements were seen for ISI. In comparison, AIR significantly improved with
the consumption of PUFA, whether in place of carbohydrate, SFA, or even MUFA (p<0.004
each).
Fig 1. Flow diagram of systematic review of published trials evaluating effects of isocaloric replacement between macronutrient
consumption on glucose homeostasis. *See S3 Text for details of the databases, eligibility criteria, search terms, and prior review articles.
doi:10.1371/journal.pmed.1002087.g001
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 7/18
Table 1. Characteristics of 102 randomised controlled feeding trials (total 239 intervention arms,
4,220 participants) evaluating effects of isocaloric replacement of dietary fats and carbohydrate on
glucose-insulin homeostasis.*
Characteristics of trials or publications nof trials or median (range)
Publication year
2000 or earlier 31
2000 to 2009 38
2010 or later 33
Geographic area
United States, Canada 35
Europe, Australia, New Zealand 57
Asia 7
Central or South America, Africa 3
Number of intervention arms
2 76
3 21
4+ 5
Design
Parallel 33
Crossover/Latin square 67
Latin square 2
Feeding duration, days 28 (3–168)
Dietary intervention*
Total energy, MJ/day 2,148 (1,000–3,466)
Carbohydrate, % energy 47.2 (5.0–65.0)
Saturated fat, % energy 9.2 (3.0–30.8)
Monounsaturated fat, % energy 13.6 (2.5–30.0)
Polyunsaturated fat, % energy 6.4 (2.0–21.4)
Protein, % energy 16.0 (10.1–33.0)
Trans fat, % .6 (.0–3.4)
Fibre, g/4.2 MJ (1,000 kcal) 13.3 (5.5–24.4)
Caloric restriction, yes 18
Provided all meals (versus partial), yes 55
Blinding of participants, yes 62
Restricted to participants with diabetes, yes 31
nof participants per trial
<25 55
25 to 49 26
50 21
Mean age of participants, years
<30 18
30 to 49.9 29
50 55
Mean body mass index of participants, kg/m
2
<25 24
25 to 29.9 45
30 33
Mean fasting glucose, mmol/L 5.4 (4.0–11.9)
Mean glycated haemoglobin, % 7.4 (4.1–11.9)
Mean weight change during follow-up, kg -0.5 (-11.8–2.7)
(Continued)
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 8/18
Exploration of Heterogeneity
For effects on fasting glucose, several sources of heterogeneity were identified (Fig 2,S3 Table).
MUFA, compared with carbohydrate, lowered fasting glucose to a greater extent in trials with
blinded participants and in trials recruiting adults with diabetes, older age, men, or higher BMI
(pheterogeneity <0.004 each). Older age and presence of diabetes also strengthened glucose-
lowering effects of PUFA (pheterogeneity <0.002 each).
Effects on fasting glucose appeared possibly smaller in trials without participant blinding,
although these differences were not statistically significant (false-discovery corrected). Replac-
ing carbohydrate with MUFA reduced fasting glucose in participant-blinded trials; but
increased fasting glucose in participant-unblinded trials (pheterogeneity <0.001). In post hoc
analyses, whether trials provided all or partial meals did not consistently influence the direction
or strength of various findings. No significant sources of heterogeneity were observed for
effects of macronutrients on fasting insulin (Fig 3).
The HbA1c-lowering effect of PUFA, compared with SFA, was significantly larger in North
American than European trials (pheterogeneity <0.0001) (S3 Table); yet despite the statistical
heterogeneity, the direction of effects was the same. No other significant sources of heterogene-
ity were observed for effects of macronutrients on HbA1c or HOMA-IR.
Sensitivity Analyses and Small Study Bias
To evaluate robustness of the main findings, we repeated meta-analyses using random effects
in five selected strata, which were significant sources of heterogeneity: trials conducted in
Western nations; trials of adults with diabetes; trials of adults without diabetes; trials providing
whole meals; and trials with blinding of meals provided (S4 Table). Findings using random
effects were generally similar, with some results having wider CIs and failing to achieve statisti-
cal significance (e.g., for HbA1c); most results being statistically significant in both fixed-effects
and random-effects models, in particular for 2 h insulin, HOMA-IR, and AIR; and rarely some
findings being significant in random-effects but not fixed-effects models. Other sensitivity
analyses also supported robustness of our main findings, including evaluating a range of
assumed between-arm correlations in crossover or Latin-square trials (S1 Fig) and altering
model covariates, imputation methods for trans fat, and restrictions on trial subtypes (S5
Table). For example, while a smaller subset of trials (31 of 102) specifically aimed to achieve
major variation in PUFA, analysis restricted to these trials showed generally similar findings,
with wider confidence intervals, as the primary analyses. We also identified little evidence for
small study bias based on influence analysis tested by linear regression (analogous to Egger’s
test: p>0.24 each) or non-parametric Wilcoxon rank tests (p>0.28 each) (S2 Fig).
Discussion
The results of this systematic review and meta-analysis of randomised controlled feeding trials
provide, to our knowledge, the most robust available evidence for the effects of dietary fats and
carbohydrate on diverse glucose-insulin metrics. We identified divergent relationships of
Table 1. (Continued)
Characteristics of trials or publications nof trials or median (range)
Overall study quality score †8.0 (4.0–10.5)
*Intervention arms and control arms combined.
†Possible range 0 to 11 (see S2 Table for details).
doi:10.1371/journal.pmed.1002087.t001
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 9/18
specific dietary fats with different measures of glucose-insulin homeostasis. For example, only
energy intake substitution with PUFA was linked to lower fasting glucose, lower HbA1c,
improve HOMA-IR, and improve insulin secretion capacity. These effects were generally seen
whether PUFA replaced carbohydrate or SFA; interestingly, insulin secretion capacity also
improved when PUFA replaced MUFA. In comparison, MUFA consumption did not appear
to significantly influence fasting glucose, compared to others macronutrients; but was seen to
reduce HbA1c and improve HOMA-IR in comparison to either carbohydrate or SFA.
Exchange of SFA for carbohydrate had little observed effects on most measures, except for
reduced fasting insulin and a borderline significant effect on C-peptide.
These findings help inform dietary guidance on macronutrients to influence metabolic
health. Currently, major organizations recommend that SFA be replaced with MUFA or
PUFA, largely to improve lipid profiles rather than glucose-insulin metrics, for the primary
and secondary prevention of diabetes [3,4]. Our investigation of trials with relatively short
average duration (28 d) suggests that consuming more unsaturated fats (MUFA, PUFA) in
Table 2. Effects of isocaloric replacements between carbohydrate (CHO), saturated fat (SFA), monounsaturated fat (MUFA), and polyunsaturated
fat (PUFA) on metrics of glucose-insulin homeostasis in randomised controlled feeding trials.*
Outcome ntrials (arms) nadults Effects (95% CI) of isocaloric replacement of 5% dietary energy
CHO CHO CHO SFA SFA MUFA
!SFA !MUFA !PUFA !MUFA !PUFA !PUFA
Glucose, mmol/L 99 (237) 4,144 0.02 0.00 -0.02 -0.02 -0.04 -0.02
(-0.01, 0.04) (-0.02, 0.02) (-0.05, 0.01) (-0.04, 0.00) (-0.07, -0.01)*(-0.05, 0.01)
2 h glucose, mmol/L†11 (29) 615 -0.04 -0.15 0.21 -0.10 0.26 0.36
(-0.39, 0.31) (-0.76, 0.47) (-0.35, 0.78) (-0.91, 0.70) (-0.34, 0.85) (-0.48, 1.20)
Haemoglobin A1c, % 23 (54) 618 0.03 -0.09 -0.11 -0.12 -0.15 -0.03
(-0.02, 0.09) (-0.12, -0.05)*** (-0.17, -0.05)*** (-0.19, -0.05)*** (-0.23, -0.06)*** (-0.09, 0.03)
Insulin, pmol/L 90 (216) 3,774 -1.1 0.1 -1.6 1.2 -0.5 -1.6
(-1.7, -0.5)** (-0.3, 0.4) (-2.8, -0.4)*(0.6, 1.8)*** (-2.0, 1.1) (-2.8, -0.5)*
2 h insulin, pmol/L†11 (28) 598 1.9 -20.3 -24.9 -22.2 -26.8 -4.6
(-19.3, 23.1) (-32.2, -8.4)** (-53.9, 4.1) (-49.1, 4.6) (-72.5, 18.9) (-33.3, 24.1)
C-peptide, nmol/L 7 (16) 175 0.03 0.02 -0.05 -0.01 -0.07 -0.06
(0.00, 0.05)*(-0.01, 0.04) (-0.11, 0.02) (-0.03, 0.01) (-0.14, -0.01)*(-0.14, 0.01)
HOMA-IR, % change 30 (76) 1,801 0.7 -2.4 -3.4 -3.1 -4.1 -1.0
(-1.6, 3.1) (-4.6, -0.3)*(-5.9, -0.8)*(-5.8, -0.4)** (-6.4, -1.6)*(-4.4, 2.6)
Insulin sensitivity
index, 10
−5
/(pmol/L)/
min‡
13 (38) 1,292 -0.10 -0.01 0.14 0.08 0.24 0.16
(-0.21, 0.02) (-0.11, 0.08) (-0.14, 0.43) (-0.01, 0.17) (-0.13, 0.61) (-0.20, 0.52)
Acute insulin
response, pmol/L/
min‡
10 (29) 1,204 -0.02 -0.03 0.49 -0.01 0.51 0.52
(-0.11, 0.07) (-0.07, 0.01) (0.17, 0.80)** (-0.08, 0.06) (0.20, 0.82)** (0.21, 0.82)**
*Values represent the pooled mean change (95% CI) for isocaloric exchange of the specified macronutrients, with the other macronutrients held constant.
All analyses adjusted for between-arm differences in protein (% energy), trans-fat (% energy), and dietary fibre (g/1000 kcal) within each trial. 1 mg/dL
glucose = 0.0555 mmol/L; 1 mU/L insulin = 6 pmol/L; HbA1 mmol/mol = (HbA1c % - 2.15)×10.929.
*p<0.05
** p<0.01
*** p<0.001.
†Oral glucose tolerance tests evaluating post-prandial glucose levels after ingestion of a test meal or drink.
‡Positive values for the insulin sensitivity index (Minimal Model) and acute insulin response, derived from intravenous infusion tests, indicate improvement
of insulin sensitivity and insulin secretion capacity, respectively.
doi:10.1371/journal.pmed.1002087.t002
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 10 / 18
place of either carbohydrate or SFA may improve HbA1C and HOMA-IR; and that focusing
on PUFA in particular may have additional benefits on insulin secretion capacity. The compar-
atively similar effects of SFA versus carbohydrate on glucose-insulin homeostasis are consistent
with their similar overall associations with both incident diabetes and cardiovascular events
[27]. Translated to foods, these finding support increased consumption of vegetable oils and
spreads, nuts, fish, and vegetables rich in unsaturated fats (e.g., avocado), in place of either ani-
mal fats or refined grains, starches, and sugars.
The magnitudes of the observed effects deserve consideration. For example, for each 5%
energy of increased MUFA or PUFA, HbA1c improved by approximately 0.1%. Based on the
relationship between HbA1c and clinical events, a 0.1% reduction would be estimated to reduce
the incidence of type 2 diabetes by 22.0% (95% CI = 15.9, 28.4%) [28] and cardiovascular dis-
eases by 6.8% (1.3, 13.0%) [29]. Such an effect could clearly be clinically meaningful, especially
given the current global pandemic of type 2 diabetes [1,2].
While both MUFA and PUFA similarly improve blood lipid profiles [9,10], their associa-
tions with clinical cardiovascular events are less similar [27]. Due to these differences, the US
Dietary Guidelines Advisory Committee concluded that strong evidence exists for cardiovascu-
lar benefits of PUFA, but limited evidence for cardiovascular benefits of MUFA [30]. Given the
similar effects of these unsaturated fats on blood lipids, the present investigation may partly
elucidate why PUFA might have greater overall cardiovascular benefits, given its additional
benefits on fasting glucose and insulin secretion capacity, key pathological markers for devel-
opment and progression of metabolic disease. The independence of these benefits, whether
Fig 2. Effects on fasting glucose of isocaloric replacements between carbohydrate (CHO), saturated fat (SFA), monounsaturated fat (MUFA), and
polyunsaturated fat (PUFA) in randomised controlled feeding trials. Values represent pooled mean effects (95% CI) of specified macronutrient
replacements, with other macronutrients held constant. *Significant heterogeneity across strata after correction for false-discovery rate (exploration of
multiple characteristics for heterogeneity). †Estimates not shown due to wide 95% CIs; see S3 Table for numeric information. 1 mg/dL = 0.0555 mmol/L.
doi:10.1371/journal.pmed.1002087.g002
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 11 / 18
PUFA replaces carbohydrate or SFA (or for insulin secretion capacity, even MUFA), is consis-
tent with growing evidence for specific cardiometabolic benefits of PUFA, regardless of the
replacement nutrient [31,32].
Biologic plausibility of these findings is supported by experimental evidence that PUFA sup-
presses oxidative stress, hepatic lipogenesis and steatosis, pancreatic lipotoxicity, and insulin
resistance [33–37]. PUFA may also help counter toxicity of tissue free fatty acids [35]; and
increase membrane fluidity, which might augment insulin sensitivity and lower risk of type 2
diabetes [38,39]. These effects have been seen with omega-6 linoleic acid, the predominant
PUFA (generally 90%+ of total PUFA), rather than only omega-3 PUFA. Meta-analyses of
omega-3 supplementation as well as dietary intakes and blood biomarker levels of omega-3
PUFA demonstrate no significant effects on fasting glucose or incident diabetes [40,41].
Together with our results, these findings suggest that metabolic benefits of PUFA relate to
omega-6 PUFA or total PUFA, and not omega-3 PUFA alone.
Compared with PUFA (consumed from a small number of vegetable oils and nuts), MUFA
derives from diverse types of foods including red meats, dairy, nuts, and vegetable oils. Cardio-
metabolic effects of these different foods vary widely [27]: red meats and especially processed
meats appear to increase risk of diabetes; milk, cheese, and yogurt appear relatively neutral or
modestly beneficial; while specific plant sources of MUFA, such as nuts and virgin olive oil,
have cardiometabolic benefits [27,42,43]. In the present investigation, most trials that sought
to increase MUFA consumption did so via increased plant sources (olive oil, canola oil, sun-
flower oil, nuts); trials that lowered MUFA generally did so by lowering animal fats (which
Fig 3. Effects on fasting insulin of isocaloric replacements between carbohydrate (CHO), saturated fat (SFA), monounsaturated fat (MUFA), and
polyunsaturated fat (PUFA) in randomised controlled feeding trials. Values represent pooled mean effects (95% CI) of specified macronutrient
replacements, with other macronutrients held constant. No significant sources of heterogeneity were detected. †Estimates not shown due to wide, 95% CIs;
see S3 Table for numeric information. 1 μIU/mL = 6 pmol/L.
doi:10.1371/journal.pmed.1002087.g003
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 12 / 18
contain both SFA and MUFA). Thus, effects of altering MUFA consumption could vary
depending on the food source. Yet, in all these foods, the MUFA molecule is identical (nearly
entirely [>95%] oleic acid), so that if effects vary by food source, it should be due to other com-
pounds in these foods (e.g., phenolics in nuts and oils; haeme iron in meats; probiotics in
yogurt), rather than different effects of plant- versus animal-origin MUFA per se.
Our findings for SFA are consistent with observed relationships with incident diabetes and
clinical cardiovascular events. Compared to the average background diet (predominantly car-
bohydrates), SFA consumption is not associated with risk of incident diabetes in long-term
cohorts [44]; nor did reduction of SFA, when replaced with carbohydrate, alter risk of incident
diabetes in the Women’s Health Initiative randomised trial [45]. Because diabetes and insulin
resistance are major risk factors for cardiovascular disease, our findings also support and help
explain meta-analyses demonstrating no association of overall SFA consumption, when com-
pared with the average background diet or total carbohydrate, with risk of coronary heart dis-
ease or stroke [30,46].
In vitro, even-chain SFA, including myristic acid (14:0) and palmitic acid (16:0), activates
pro-inflammatory cascades, induces skeletal muscle insulin resistance, and damages pancreatic
β-cells, while the MUFA oleic acid (18:1) may partly protect against some of these effects
[35,47–49]. However, in vivo, dietary SFA and MUFA may be readily oxidized as energy
sources [50,51], while tissue levels of major SFA and MUFA may be at least equally influenced
by endogenous hepatic synthesis of fatty acids rather than direct dietary intake [52]. This
explains why dietary starch and sugars, which activate hepatic de novo lipogenesis, are posi-
tively associated with blood levels of major SFA and MUFA [52–54]. Thus, effects of blood and
tissue SFA and MUFA may not inform and should be separately considered from biologic
effects of dietary SFA and MUFA.
In exploratory analyses, we identified some sources of potential heterogeneity in effects of
dietary macronutrients. The most compelling interactions, based on consistency across differ-
ent measures and with reasonably large numbers of trials in each subgroup, were for stronger
benefits of MUFA and PUFA on fasting glucose among older adults and patients with preva-
lent diabetes. Both our identified and null findings for heterogeneity should be interpreted with
caution: absence of significant heterogeneity could result from insufficient power (e.g., by
region, trials in non-Western countries were scarce), while positive interaction could result
from chance, even corrected for false-discovery. Our findings advance the field by exploring
interactions using all currently available data from feeding trials, which generate hypotheses to
be tested in new studies, including studies of gene-diet interactions across diverse populations,
controlled trials of glucose-insulin biomarkers, and prospective studies of clinical events.
Our investigation has several strengths. Our systematic search, rigorous screening, and data
extraction protocols made it unlikely that any large studies or relevant data were missed or erro-
neously extracted. In addition, the large number of identified studies makes it unlikely that any
single study, whether included or missed, would appreciably alter our findings. We focused on
randomised, controlled trials using feeding interventions, maximizing inference for true biologi-
cal effects. We examined different replacement scenarios among major macronutrients, provid-
ing novel insights for the most relevant replacements; confirmed robustness of our findings in
sensitivity analyses and adjusted for between-arm differences in protein, trans fat, and dietary
fibre, reducing the influence of variation in these factors. We evaluated multiple relevant metrics,
including fasting, post-prandial, and long-term glycaemia, insulin levels, and insulin resistance,
providing a more comprehensive picture of the full effects of dietary macronutrients.
Potential limitations should be considered. While feeding trials maximize inference for bio-
logic effects, the findings may not be generalisable to effects of dietary advice, which can be
influenced by knowledge and compliance, and to effects of long-term habitual diet. Conversely,
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 13 / 18
we found little evidence for heterogeneity by duration of intervention ranging from 3 to 168 d,
and our overall findings are consistent with meta-analyses of incident diabetes and clinical car-
diovascular events. While all trials were randomised, not all were double blind; yet, food-based
dietary trials are often, by necessity, challenging to blind for participants. This importance was
implicated in our study because replacing SFA or carbohydrate with MUFA was shown to
lower fasting glucose, 2 h glucose, 2 h insulin and HOMA-IR in trials implementing blinding
intervention but not in trials not blinding for participants. Sufficient information was not avail-
able to classify subtypes of fatty acids, so our findings should be considered most relevant to
effects of total dietary SFA (predominantly palmitic acid), total PUFA (predominantly linoleic
acid), total MUFA (almost entirely oleic acid), and total carbohydrate (mostly refined starch
and sugars). For instance, our results should not be extrapolated to potential effects of carbohy-
drate in fruit, legumes, or minimally processed whole grains. Trials inconsistently provided
information on food sources of macronutrients (e.g., specific oils) or cooking methods; future
studies should evaluate whether these characteristics modify physiologic effects. Most trials
were in North America and Europe, and findings may not be generalisable to other world
regions. Our analysis evaluated relatively few trials measuring C-peptide, post-challenge glu-
cose and insulin, ISI, and AIR, and did not evaluate outcomes specific to peripheral or hepatic
insulin sensitivity, not capturing the potential effects of fatty acids on insulin sensitivity of spe-
cific tissues. Unmeasured sources of heterogeneity may exist, such as effects of genes and cook-
ing methods. Therefore, our meta-analysis highlights the gaps in knowledge for potential
effect-modifiers for various metrics of glucose-insulin homeostasis. Our results and available
evidence support the importance of further experimental studies and large, adequately powered
feeding trials examining ISI and AIR. Meta-analyses can be influenced by small study bias; yet,
influence analysis did not support the presence of such bias, and findings for our main end-
points were based on large numbers of trials, making it unlikely that inclusion of any unpub-
lished trials would substantially alter the results.
In conclusion, this systematic review and meta-analysis provides novel quantitative evi-
dence for effects of major dietary fats and carbohydrate on glucose-insulin homeostasis. The
results support guidelines to increase MUFA intake to improve glycaemia and insulin resis-
tance, with possibly stronger effects among patients with type 2 diabetes, and to increase PUFA
intake in the general population to improve long-term glycaemic control, insulin resistance,
and insulin secretion capacity, in place of SFA or carbohydrate. These findings help inform
public health and clinical dietary guidelines to improve metabolic health.
Supporting Information
S1 Fig. Effects of isocaloric macronutrient exchange by 5% of total energy intake on (A) fast-
ing glucose, (B) haemoglobin A1c, and (C) fasting insulin under different assumption of a
between-arm correlation in crossover or Latin-square trials.
(PDF)
S2 Fig. Assessment for small study bias in meta-regression using influence analysis, evaluating
effects of isocaloric exchange of 5% energy between different macronutrients on (A) fasting
glucose, (B) haemoglobin A1c, and (C) fasting insulin.
(PDF)
S1 Table. Characteristics of 102 randomised controlled feeding trials evaluated in the
meta-analysis of effects of diets with different macronutrient compositions on glucose-
insulin homeostasis.
(PDF)
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 14 / 18
S2 Table. Characteristics and scores of reporting quality of 102 trials eligible for the meta-
analysis of randomised controlled feeding trials of macronutrient intakes and glycaemic
outcomes.
(PDF)
S3 Table. Effects of isocalorically exchanging 5% of dietary energy between carbohydrate
and major dietary fats on glucose-insulin metrics, with stratification by country, age, sex,
diabetes status, provision of meals, and blinding in randomised controlled feeding trials.
(PDF)
S4 Table. Effects of isocalorically exchanging 5% of dietary energy between carbohydrate
and major dietary fats on glucose-insulin metrics: fixed-effects and random-effects meta-
analyses by region, diabetes status, provision of meals, and blinding in randomised con-
trolled feeding trials.
(PDF)
S5 Table. Effects of isocalorically exchanging 5% of dietary energy between carbohydrate
and major dietary fats on fasting glucose, haemoglobin A1c, and fasting insulin: sensitivity
meta-analysis concerning model covariates and study characteristics.
(PDF)
S1 Text. Protocol of a systematic review and meta-analysis of effects of macronutrient
replacement on glucose-insulin homeostasis.
(PDF)
S2 Text. PRISMA 2009 Checklist.
(PDF)
S3 Text. Eligibility criteria, literature search, data preparation, imputation, and reference list.
(PDF)
Acknowledgments
We acknowledge the following individuals for provision of additional information: Dr. Marie
P. St-Onge at New York Obesity Research Center, Columbia University College of Physicians
and Surgeons, New York, United States; Prof. Parveen Yaqoob at Hugh Sinclair Unit of Nutri-
tion, School of Food Biosciences, The University of Reading, Whiteknights, United Kingdom;
Dr. Leo Niskanen, Departments of Clinical Nutrition and Internal Medicine, University of Kuo-
pio, the Finnish Red Cross Blood Transfusion Service, Helsinki, Finland; Dr. David Iggman and
Dr. Ulf Risérus, Clinical Nutrition and Metabolism, Department of Public Health and Caring Sci-
ences, Uppsala University, Uppsala, Sweden; Dr. Mary Gannon, Minneapolis VA Health Care
System, Department of Food Science and Nutrition, University of Minnesota, Minnesota, United
States; Dr. Ursula Schwab and Dr. Matti Uusitupa, Department of Clinical Nutrition, Institute of
Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finland; Dr. Jeremy
D. Furtado, Department of Nutrition, Harvard T. H. Chan School of Public Health, Massachu-
setts, United States; Prof. Barbara A Gower, Department of Nutrition Sciences, Division of Physi-
ology and Metabolism, University of Alabama at Birmingham School of Medicine, Alabama,
United States; Dr. Mohammad Javad Hosseinzadeh-Attar, Department of Clinical Nutrition,
School of Nutritional Sciences and Diabetics, Tehran University of Medical Sciences; Prof.
Manny Noakes and Dr. Tom Wycherley, the Commonwealth Scientific and Industrial Research
Organisation; University of South Australia School of Health Sciences; and Prof. Ronald M. Kar-
uss and Dr Sally Chiu, Children’s Hospital Oakland Research Institute.
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 15 / 18
Author Contributions
Conceived and designed the experiments: FI RM JHYW DM. Analyzed the data: FI. Contrib-
uted reagents/materials/analysis tools: FI RM JHYW MCdOO FOO AIA. Wrote the first draft
of the manuscript: FI. Contributed to the writing of the manuscript: FI DM. Agree with the
manuscript’s results and conclusions: FI RM JHYW MCdOO FOO AIA DM. All authors have
read, and confirm that they meet, ICMJE criteria for authorship.
References
1. NCD Risk Factor Collaboration. Worldwide trends in diabetes since 1980: a pooled analysis of 751 pop-
ulation-based studies with 44 million participants. Lancet. 2016; 387(10027):1513–30. doi: 10.1016/
S0140-6736(16)00618-8 PMID: 27061677
2. International Diabetes Federatoin. The Global Burden. In: IDF Diabetes Atlas. 6th ed. Brussels, Bel-
gium: International Diabetes Federation; 2013. p. 29–49.
3. Deakin T, Duncan A, Dyson P, Frost G, Harrison Z, Kelly T, et al. Evidence-based nutrition guidelines
for the prevention and management of diabetes. Kelly T, Dyson P, editors. 2011. Available:https://www.
diabetes.org.uk/About_us/What-we-say/Food-nutrition-lifestyle/Evidence-based-nutrition-guidelines-
for-the-prevention-and-management-of-diabetes-May-2011/ (accessed 2016 Apr 29)
4. Evert AB, Boucher JL, Cypress M, Dunbar SA, Franz MJ, Mayer-Davis EJ, et al. Nutrition therapy rec-
ommendations for the management of adults with diabetes. Diabetes Care. 2013; 36(11):3821–42. doi:
10.2337/dc13-2042 PMID: 24107659
5. Fats and fatty acids in human nutrition. Roma, Italy: Food and Agriculture Organization; 2010.
6. Aranceta J, Pérez-Rodrigo C. Recommended dietary reference intakes, nutritional goals and dietary
guidelines for fat and fatty acids: a systematic review. Br J Nutr. 2012; 107(Suppl 2):S8–22. doi: 10.
1017/S0007114512001444 PMID: 22591906
7. Eckel RH, Jakicic JM, Ard JD, de Jesus JM, Houston Miller N, Hubbard VS, et al. 2013 AHA/ACC
guideline on lifestyle management to reduce cardiovascular risk: a report of the American College of
Cardiology/American Heart Association Task Force on Practice Guidelines. Circulation. 2014; 129(25
Suppl 2):S76–99. doi: 10.1161/01.cir.0000437740.48606.d1 PMID: 24222015
8. Galgani J, Uauy R, Aguirre CA, Díaz EO. Effect of the dietary fat quality on insulin sensitivity. Br J Nutr.
2008; 100(3):471–9. doi: 10.1017/S0007114508894408 PMID: 18394213
9. Mensink RP, Zock PL, Kester ADM, Katan MB. Effects of dietary fatty acids and carbohydrates on the
ratio of serum total to HDL cholesterol and on serum lipids and apolipoproteins: a meta-analysis of 60
controlled trials. Am J Clin Nutr. 2003; 77(5):1146–55. PMID: 12716665
10. Clarke R, Frost C, Collins R, Appleby P, Peto R. Dietary lipids and blood cholesterol: quantitative meta-
analysis of metabolic ward studies. BMJ. 1997; 314(7074):112. doi: 10.1136/bmj.314.7074.112 PMID:
9006469
11. Willett WC. Implications of Total Energy Intake for Epidemiologic Analyses. In: Willett WC, editor. Nutri-
tional Epidemiology. 3rd ed. New York, USA: Oxford University Press; 2012. p. 260–86.
12. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JPA, et al. The PRISMA statement
for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions:
explanation and elaboration. PLoS Med. 2009; 6(7):e1000100. doi: 10.1371/journal.pmed.1000100
PMID: 19621070
13. Wallace TM, Matthews DR. The assessment of insulin resistance in man. Diabet Med. 2002; 19
(7):527–34. doi: 10.1046/j.1464-5491.2002.00745.x PMID: 12099954
14. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJM, Gavaghan DJ, et al. Assessing the qual-
ity of reports of randomized clinical trials: Is blinding necessary? Control Clin Trials. 1996; 17(1):1–12.
doi: 10.1016/0197-2456(95)00134-4 PMID: 8721797
15. Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions. 1st ed. Higgins J,
Green S, editors. West Sussex, England: The Cochrane Collaboration; 2008.
16. Elbourne DR. Meta-analyses involving cross-over trials: methodological issues. International Journal of
Epidemiology. 2002; 31(1):140–9. doi: 10.1093/ije/31.1.140 PMID: 11914310
17. Furukawa TA, Barbui C, Cipriani A, Brambilla P, Watanabe N. Imputing missing standard deviations in
meta-analyses can provide accurate results. J Clin Epidemiol. 2006; 59(1):7–10. doi: 10.1016/j.jclinepi.
2005.06.006 PMID: 16360555
18. White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and guidance for
practice. Stat Med. 2011; 30(4):377–99. doi: 10.1002/sim.4067 PMID: 21225900
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 16 / 18
19. The GLIMMIX Procedure. In: SAS/STAT(R) 93 User’s Guide. North Carolina, US: SAS Institute Inc.;
2011.
20. Salanti G, Higgins JPT, Ades AE, Ioannidis JPA. Evaluation of networks of randomized trials. Stat
Methods Med Res. 2008; 17(3):279–301. doi: 10.1177/0962280207080643 PMID: 17925316
21. Higgins JP, Altman DG, Gotzsche PC, Juni P, Moher D, Oxman AD, et al. The Cochrane Collabora-
tion’s tool for assessing risk of bias in randomised trials. BMJ. 2011; 343:d5928. doi: 10.1136/bmj.
d5928 PMID: 22008217
22. Yekutieli D, Benjamini Y. The control of the false discovery rate in multiple testing under dependency.
Ann Stat. 2001; 29(4):1165–88. doi: 10.1214/aos/1013699998
23. Wycherley TP, Moran LJ, Clifton PM, Noakes M, Brinkworth GD. Effects of energy-restricted high-pro-
tein, low-fat compared with standard-protein, low-fat diets: a meta-analysis of randomized controlled tri-
als. Am J Clin Nutr. 2012; 96(6):1281–98. doi: 10.3945/ajcn.112.044321 PMID: 23097268
24. Ebbeling CB, Swain JF, Feldman H a, Wong WW, Hachey DL, Garcia-Lago E, et al. Effects of dietary
composition on energy expenditure during weight-loss maintenance. JAMA. 2012; 307(24):2627–34.
doi: 10.1001/jama.2012.6607 PMID: 22735432
25. Tahara Y. On the Weighted-Average Relationship Between Plasma Glucose and HbA1c: Response to
Trevino. Diabetes Care. 2006; 29(2):466–7. doi: 10.2337/diacare.29.02.06.dc05-2104 PMID:
16443914
26. Hanley JA, Negassa A, Edwardes MD deB., Forrester JE. Statistical Analysis ofCorrelated Data Using
Generalized Estimating Equations: An Orientation. Am J Epidemiol. 2003; 157(4):364–75. doi: 10.
1093/aje/kwf215 PMID: 12578807
27. Mozaffarian D. Dietary and Policy Priorities for Cardiovascular Disease, Diabetes, and Obesity: A Com-
prehensive Review. Circulation. 2016;CIRCULATIONAHA.115.018585. doi: 10.1161/
CIRCULATIONAHA.115.018585
28. Chamnan P, Simmons RK, Forouhi NG, Luben RN, Khaw K-T, Wareham NJ, et al. Incidence of Type 2
Diabetes Using Proposed HbA1c Diagnostic Criteria in the European Prospective Investigation of Can-
cer-Norfolk Cohort: Implications for preventive strategies. Diabetes Care. 2011; 34(4):950–6. doi: 10.
2337/dc09-2326 PMID: 20622160
29. Di Angelantonio E, Gao P, Khan H, Butterworth AS, Wormser D, Kaptoge S, et al. Glycated Hemoglo-
bin Measurement and Prediction of Cardiovascular Disease. JAMA. 2014; 311(12):1225. doi: 10.1001/
jama.2014.1873 PMID: 24668104
30. The 2015 Dietary Guidelines Advisory Committee. Scientific Report of the 2015 Dietary Guidelines
Advisory Committee. 2015. Available:http://health.gov/dietaryguidelines/2015-scientific-report/
(accessed 2016 Apr 28)
31. Farvid MS, Ding M, Pan A, Sun Q, Chiuve SE, Steffen LM, et al. Dietary Linoleic Acid and Risk of Coro-
nary Heart Disease: A Systematic Review and Meta-Analysis of Prospective Cohort Studies. Circula-
tion. 2014; 130(18):1568–78. doi: 10.1161/CIRCULATIONAHA.114.010236 PMID: 25161045
32. Risérus U, Willett WC, Hu FB, Riserus U, Willett WC, Hu FB. Dietary fats and prevention of type 2 dia-
betes. Prog Lipid Res. 2009; 48(1):44–51. doi: 10.1016/j.plipres.2008.10.002 PMID: 19032965
33. Jump DB. Dietary polyunsaturated fatty acids and regulation of gene transcription. Curr Opin Lipidol.
2002; 13(2):155–64. doi: 10.1097/00041433-200204000-00007 PMID: 11891418
34. Giordano E, Visioli F. Long-chain omega 3 fatty acids: molecular bases of potential antioxidant actions.
Prostaglandins Leukot Essent Fatty Acids. 2014; 90(1):1–4. doi: 10.1016/j.plefa.2013.11.002 PMID:
24345866
35. Cao H, Gerhold K, Mayers JR, Wiest MM, Watkins SM, Hotamisligil GS. Identification of a lipokine, a
lipid hormone linking adipose tissue to systemic metabolism. Cell. 2008; 134(6):933–44. doi: 10.1016/j.
cell.2008.07.048 PMID: 18805087
36. Bjermo H, Iggman D, Kullberg J, Dahlman I, Johansson L, Persson L, et al. Effects of n-6 PUFAs com-
pared with SFAs on liver fat, lipoproteins, and inflammation in abdominal obesity: a randomized con-
trolled trial. Am J Clin Nutr. 2012; 95(5):1003–12. doi: 10.3945/ajcn.111.030114 PMID: 22492369
37. Rosqvist F, Iggman D, Kullberg J, Cedernaes J, Johansson HE, Larsson A, et al. Overfeeding Polyun-
saturated and Saturated Fat Causes Distinct Effects on Liver and Visceral Fat Accumulation in
Humans. Diabetes. 2014; 63(7):2356–68. doi: 10.2337/db13-1622 PMID: 24550191
38. Kröger J, Jacobs S, Jansen EHJM, Fritsche A, Boeing H, Schulze MB. Erythrocyte membrane fatty
acid fluidity and risk of type 2 diabetes in the EPIC-Potsdam study. Diabetologia. 2015; 58(2):282–9.
doi: 10.1007/s00125-014-3421-7 PMID: 25344391
39. Dutta-Roy AK. Insulin mediated processes in platelets, erythrocytes and monocytes/macrophages:
Effects of essential fatty acid metabolism. Prostaglandins, Leukotrienes and Essential Fatty Acids.
1994; 51(6):385–99. doi: 10.1016/0952-3278(94)90054-X
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 17 / 18
40. Hartweg J, Perera R, Montori V, Dinneen S, Neil HA, Farmer A. Omega-3 polyunsaturated fatty acids
(PUFA) for type 2 diabetes mellitus. Cochrane Database Syst Rev. 2008;(1: ):CD003205. doi: 10.1002/
14651858.CD003205.pub2 PMID: 18254017
41. Wu JHY, Micha R, Imamura F, Pan A, Biggs ML, Ajaz O, et al. Omega-3 fatty acids and incident type 2
diabetes: a systematic review and meta-analysis. Br J Nutr. 2012; 107 Suppl:S214–27. doi: 10.1017/
S0007114512001602
42. Afshin A, Micha R, Khatibzadeh S, Mozaffarian D. Consumption of nuts and legumes and risk of inci-
dent ischemic heart disease, stroke, and diabetes: a systematic review and meta-analysis. Am J Clin
Nutr. 2014; 100(1):278–88. doi: 10.3945/ajcn.113.076901 PMID: 24898241
43. Martínez-González MA, Salas-Salvadó J, Estruch R, Corella D D, Fitó M, Ros E. Benefits of the Medi-
terranean Diet: Insights from the PREDIMED Study. Prog Cardiovasc Dis. 2015; 58(1):50–60. doi: 10.
1016/j.pcad.2015.04.003 PMID: 25940230
44. Micha R, Mozaffarian D. Saturated Fat and Cardiometabolic Risk Factors,Coronary Heart Disease,
Stroke, and Diabetes: a Fresh Look at the Evidence. Lipids. 2010; 45(10):893–905. doi: 10.1007/
s11745-010-3393-4 PMID: 20354806
45. Tinker LF, Bonds DE, Margolis KL, Manson JE, Howard B V, Larson J, et al. Low-fat dietary pattern and
risk of treated diabetes mellitus in postmenopausal women: the Women’s Health Initiative randomized
controlled dietary modification trial. Arch Intern Med. 2008; 168(14):1500–11. doi: 10.1001/archinte.
168.14.1500 PMID: 18663162
46. Chowdhury R, Warnakula S, Kunutsor S, Crowe F, Ward HA, Johnson L, et al. Association of Dietary,
Circulating, and Supplement Fatty Acids With Coronary Risk. Ann Int Med. 2014; 160(6):398–406. doi:
10.7326/M13-1788 PMID: 24723079
47. Maedler K, Oberholzer J, Bucher P, Spinas GA, Donath MY. Monounsaturated Fatty Acids Prevent the
Deleterious Effects of Palmitate and High Glucose on Human Pancreatic -Cell Turnover and Function.
Diabetes. 2003; 52(3):726–33. doi: 10.2337/diabetes.52.3.726 PMID: 12606514
48. Morgan NG, Dhayal S, Diakogiannaki E, Welters HJ. The cytoprotective actions of long-chain mono-
unsaturated fatty acids in pancreatic beta-cells. Biochem Soc Trans. 2008; 36(Pt 5):905–8. doi: 10.
1042/BST0360905 PMID: 18793159
49. Silveira LR, Fiamoncini J, Hirabara SM, Procópio J, Cambiaghi TD, Pinheiro CHJ, et al. Updating the
effects of fatty acids on skeletal muscle. J Cell Physiol. 2008; 217(1):1–12. doi: 10.1002/jcp.21514
PMID: 18543263
50. Schwarz J-M, Linfoot P, Dare D, Aghajanian K. Hepatic de novo lipogenesis in normoinsulinemic and
hyperinsulinemic subjects consuming high-fat, low-carbohydrate and low-fat, high-carbohydrate isoe-
nergetic diets. Am J Clin Nutr. 2003; 77(1):43–50. PMID: 12499321
51. Roberts R, Bickerton AS, Fielding BA, Blaak EE, Wagenmakers AJ, Chong MFF, et al. Reduced oxida-
tion of dietary fat after a short term high-carbohydrate diet. Am J Clin Nutr. 2008; 87(4):824–31. PMID:
18400703
52. Flowers MT, Ntambi JM. Stearoyl-CoA desaturase and its relation to high-carbohydrate diets and obe-
sity. Biochim Biophys Acta. 2009; 1791(2):85–91. doi: 10.1016/j.bbalip.2008.12.011 PMID: 19166967
53. Forouhi NG, Koulman A, Sharp SJ, Imamura F, Kröger J, Schulze MB, et al. Differences in the prospec-
tive association between individual plasma phospholipid saturated fatty acids and incident type 2 diabe-
tes: the EPIC-InterAct case-cohort study. Lancet Diabetes Endocrinol. 2014; 2(10):810–8. doi: 10.
1016/S2213-8587(14)70146-9 PMID: 25107467
54. Volk BM, Kunces LJ, Freidenreich DJ, Kupchak BR, Saenz C, Artistizabal JC, et al. Effects of Step-
Wise Increases in Dietary Carbohydrate on Circulating Saturated Fatty Acids and Palmitoleic Acid in
Adults with Metabolic Syndrome. PLoS ONE. 2014; 9(11):e113605. doi: 10.1371/journal.pone.
0113605 PMID: 25415333
Macronutrients and Glucose-Insulin Homeostasis
PLOS Medicine | DOI:10.1371/journal.pmed.1002087 July 19, 2016 18 / 18
