caffeinated and decaffeinated coffee consumption and risk of … · health implications. an inverse...
TRANSCRIPT
Caffeinated and DecaffeinatedCoffee Consumption and Riskof Type 2 Diabetes: A SystematicReview and a Dose-ResponseMeta-analysis
OBJECTIVE
Previous meta-analyses identified an inverse association of coffee consumptionwith the risk of type 2 diabetes. However, an updated meta-analysis is neededbecause new studies comparing the trends of association for caffeinated anddecaffeinated coffee have since been published.
RESEARCH DESIGN AND METHODS
PubMed and Embase were searched for cohort or nested case-control studiesthat assessed the relationship of coffee consumption and risk of type 2 diabetesfrom1966 to February 2013. A restricted cubic spline random-effectsmodelwasused.
RESULTS
Twenty-eight prospective studies were included in the analysis, with 1,109,272study participants and 45,335 cases of type 2 diabetes. The follow-up durationranged from 10 months to 20 years. Compared with no or rare coffee consump-tion, the relative risk (RR; 95% CI) for diabetes was 0.92 (0.90–0.94), 0.85 (0.82–0.88), 0.79 (0.75–0.83), 0.75 (0.71–0.80), 0.71 (0.65–0.76), and 0.67 (0.61–0.74) for1–6 cups/day, respectively. The RR of diabetes for a 1 cup/day increase was 0.91(0.89–0.94) for caffeinated coffee consumption and 0.94 (0.91–0.98) for decaf-feinated coffee consumption (P for difference = 0.17).
CONCLUSIONS
Coffee consumption was inversely associated with the risk of type 2 diabetes in adose-responsemanner. Both caffeinated and decaffeinated coffeewas associatedwith reduced diabetes risk.Diabetes Care 2014;37:569–586 | DOI: 10.2337/dc13-1203
Type 2 diabetes is a chronic disease with high rates of morbidity and mortality. Theworldwide prevalence of type 2 diabetes is increasing, and the global number ofpeople with diabetes is estimated to reach 366 million by the year 2030 (1). The riskof blindness, renal disease, and amputation among those with type 2 diabetes is 20to 40 times higher than that of people without diabetes. Furthermore, those withtype 2 diabetes have a two to five times higher risk of myocardial infarction and twoto three times higher risk of stroke (2). Given its significant burden, identifyingmodifiable lifestyle factors is imperative for the prevention of diabetes.
1Department of Nutrition, Harvard School ofPublic Health, Boston, MA2Saw Swee Hock School of Public Health andDepartment of Medicine, Yong Loo Lin School ofMedicine, National University of Singapore andNational University Health System, Singapore3Department of Epidemiology, Harvard School ofPublic Health, Boston, MA4Channing Division of Network Medicine,Brigham and Women’s Hospital and HarvardMedical School, Boston, MA
Corresponding author: Frank B. Hu, [email protected].
Received 22 May 2013 and accepted 27September 2013.
This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc13-1203/-/DC1.
© 2014 by the American Diabetes Association.See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.
Ming Ding,1 Shilpa N. Bhupathiraju,1
Mu Chen,1 Rob M. van Dam,1,2
and Frank B. Hu1,3,4
Diabetes Care Volume 37, February 2014 569
META
-ANALYSIS
Coffee is one of the most widelyconsumed beverages around the world;thus, investigating its association withvarious diseases has important publichealth implications. An inverseassociation between coffee consumptionand risk of type 2 diabetes was firstreported in a Dutch population (3). Insubsequent years, this finding has beenconfirmed in most (4–10) but not all (11–13) studies. Two previous meta-analysesof coffee consumption and risk of type 2diabetes have been published. van Damand Hu (14) included nine cohort studiesand reported a lower risk of type 2diabetes for high coffee consumptioncompared with no coffee consumptionwith little between-study heterogeneity.Huxley et al. (15) included 18 prospectivestudies and found an inversemonotonousassociation between the number of cupsof coffee consumed and diabetes risk, butthey double-counted data from the samecohort (16,17). Since the publication ofthese meta-analyses, 10 additionalprospective studies on the associationbetween coffee consumption anddiabetes have been published(6,8,10,12,18–23).
A key issue that remains to be resolved iswhether consumption of caffeinated anddecaffeinated coffee is similarlyassociatedwith the risk of type 2 diabetes.Such results would provide insight intothe role of caffeine in the putativerelationship between coffee consumptionand diabetes risk. The meta-analysis byHuxley et al. (15) included only six studieson decaffeinated coffee, and estimateswere less precise than for caffeinatedcoffee because of lower consumptionlevels. Eight subsequent prospectivestudies evaluated the association withdecaffeinated coffee, approximatelydoubling the amount of data ondecaffeinated coffee and diabetes risk(8,10,12,18–21,24). We thereforeperformed an updated systematic reviewand a dose-response meta-analysis of allavailable data on the association of bothcaffeinated and decaffeinated coffeeconsumption with the risk of type 2diabetes.
RESEARCH DESIGN AND METHODS
Search Strategy and Selection CriteriaWe searched the PubMed and Embasedatabases for prospective cohort
studies or nested case-control studiesthat evaluated the association betweencoffee consumption and risk of type 2diabetes between January 1966 andFebruary 2013. The computer-basedsearches included the key words coffeeand diabetes. No Medical SubjectHeadings terms were used because ofthe clear definitions of coffee anddiabetes. Reference lists of retrievedarticles were manually scanned for allrelevant additional studies and reviewarticles. We restricted the search tostudies on humans and written in English.
Study SelectionStudies included in this meta-analysismet the following criteria: 1) the studydesign was prospective cohort or nestedcase-control; 2) the articles werepublished in English; 3) the exposurewas categorized coffee consumption,including total coffee, caffeinatedcoffee, or decaffeinated coffee; and 4)the outcome was risk of type 2 diabetes.Studies were excluded if they werecross-sectional in design and ifinformation on dose-response modelingwas inadequate, including the numberof participants and cases, relative risk(RR) and SE for the estimate, and dose ofcoffee consumption in each exposurecategory.
Data Extraction and QualityAssessmentOne author (M.D.) assessed studyeligibility and extracted the data, andanother (M.C.) independently double-checked the available data. Thefollowing data were extracted fromeach study: first author’s name, year ofpublication, geographical location,follow-up time, sex, age, number of type2 diabetes events, number ofparticipants, categories of coffeeconsumption, type of assessment oftype 2 diabetes, covariates adjusted forin themultivariable analysis, and RRs forall categories of coffee consumption. Ifstudies based on the same cohort werepublished several times, only the mostrecent article was selected. If a studyonly reported caffeinated coffeeconsumption instead of total coffeeconsumption, caffeinated coffeeconsumption was also included in thetotal coffee consumption analysis. TheNewcastle-Ottawa quality assessment
scale (NOS) was used to evaluate thestudy quality of the included articles.The scale ranges from 0 to 9 points, withhigher points indicating higher studyquality. Specifically, loss to follow-upwas assessed as an item in the NOS forindividual studies. In the present meta-analysis, we assumed that loss to follow-up was random.
To perform a dose-response meta-analysis, we assigned the median ormean coffee consumption in eachcategory of consumption to thecorresponding RR for each study. If themean or median consumption percategory was not reported, themidpoint of the upper and lowerboundaries in each category wasassigned as the mean consumption. Ifthe upper boundary for the highestcategory was not provided, the assignedmean value was 25% higher than thelower boundary of that category. If thelower boundary for the lowest categorywas not provided, the assigned meanvalue was one-half of the upperboundary of that category.
Data Synthesis and AnalysisTo analyze the trend of coffeeconsumption and risk of type 2 diabetes,we used both semiparametric andparametric methods. For thesemiparametric method, four coffeeconsumption groups were generated,namely lowest, third highest, secondhighest, and highest. For each includedstudy, the lowest and the highest coffeeconsumption categories correspondedto the lowest and highest groups,respectively. For studies with fourexposure categories, the second andthird categories corresponded to thesecond and third highest groups,respectively. For studies with threeexposure categories, the middlecategory corresponded to either thesecond or the third highest group,whichever median coffee consumptionamount was most similar. If the studyhad more than four exposurecategories, two consumption groups,other than the lowest and highest, werechosen on the basis of their similarity ofthe amount of coffee consumption inthat category to the second and thirdhighest groups. For each group, arandom-effects model was used to poolthe RR of type 2 diabetes (25). Statistical
570 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
heterogeneity in each group wasassessed by Cochrane Q test, with P ,0.1 indicating significant between-studyheterogeneity.
For the parametric method, a random-effects dose-response meta-analysiswas performed (26). The number ofcases and participants in each coffeeconsumption category was extracted toestimate the covariance of the RR ineach study. Together with the observedadjusted variance of the RR, weestimated the variance/covariancematrix of the data. The weight of eachstudy was calculated as the inverse ofthe variance/covariance matrix. Weused generalized least squares modelswith maximum likelihood method toestimate the coefficients for eachstudy, and we fit a random-effectsgeneralized linear model. Additionally,we tested for potential nonlinearityin the association between coffeeconsumption and type 2 diabetes by arandom-effects restricted cubic splinemodel with three knots. Statacommands GLST for model fitting andLINCOME for estimating effect wereused for the fitted model.
Stratified analyses were performed forstudy location, sex, diabetes assessment
method, and type of coffee (i.e.,caffeinated, decaffeinated), and thesubgroup difference was tested in theassociation of coffee consumption withrisk of type 2 diabetes. We assessed thepotential for publication bias through aformal Egger regression symmetry test.All analyses were conducted with Stata10 (StataCorp, College Station, TX)statistical software.
RESULTS
Characteristics of StudiesThe initial search identified 800potentially relevant citations. Afterscreening titles and abstracts, weidentified 39 studies for furtherevaluation (Fig. 1). We excluded sixstudies with a cross-sectional studydesign, three that were abstracts ofconference posters, and two that wererepeated in the same cohort. The resultsof the remaining 28 studies, comprising1,109,272 study participants and 45,335cases of type 2 diabetes, were includedin the meta-analysis (3–13,17–24,27–35). Characteristics of all 28 studieswere shown in Table 1. The duration offollow-up for incident type 2 diabetesranged from 10 months to 20 years,with a median follow-up of 11 years.
Thirteen studies were conducted in theU.S., 11 in Europe, and 4 in Asia. In ninestudies, type 2 diabetes was self-reported; the outcome of six studieswasassessed by means of a glucosetolerance test, and the outcome of theother studies was confirmed by eithermedical records or national registries.One study was a nested case-controldesign, and the remaining 27 wereprospective cohort studies. Ten studiesassessed both caffeinated anddecaffeinated coffee, and only one eachassessed caffeinated coffee anddecaffeinated coffee. Fifteen studies didnot distinguish caffeinated coffee anddecaffeinated coffee, and sevenassessed caffeine consumption. Themean NOS score was 7 (of a possible 9points), suggesting a high quality of thestudies included in the meta-analysis.The points were mainly lost in exposureassessment and adequacy of follow-upof cohorts: Six studies assessed coffeeconsumption by structured interview,and five addressed the percentage ofloss to follow-up.
Total Coffee Consumption and Risk ofType 2 DiabetesFigure 2 shows the results of differentlevels of total coffee consumption
Figure 1—Study selection process of the identified articles.
care.diabetesjournals.org Ding and Associates 571
Table
1—Basicch
aracteristicsoftheincludedstudies
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
vanDam
andFesken
s20
02(3)
Europe
Both
30–60
1–13
17,111
306
Self-rep
orted
Totalcoffee
Educationallevel,leisure
timephysicalactivity,
occup
ationalp
hysical
activity,alcoho
l
consum
ption,cigarette
smoking,history
of
cardiovascular
disease,
know
nhypertension
,
andknow
n
hypercholesterolemia
6
#2
1.00(1.00–
1.00
)
3–4
0.79(0.57–
1.10
)
5–6
0.73(0.53–
1.01
)
$7
0.50(0.35–
0.72
)
Reu
nan
enet
al.20
03(11)
Europe
Both
20–98
304,227
a19
,518
855
Confirm
edcases
Totalcoffee
Age,sex,BMI,sm
oking,
andleisure
time
physicalactivity
8
#2
1.00(1.00–
1.00
)
3–4
1.01(0.81–
1.27
)
5–6
0.98(0.79–
1.21
)
$7
0.92(0.73–
1.16
)
Carlssonet
al.20
04(7)
Europe
Both
30–60
2010
,652
408
Confirm
edcases
Totalcoffee
Age,sex,BMI,ed
ucation,
leisure
timephysical
activity,alcohol
consumption,and
smoking
6
#2
1.00(1.00–
1.00
)
3–4
0.70(0.48–
1.01
)
5–6
0.71(0.50–
1.01
)
$7
0.65(0.44–
0.96
)
Rosengren
etal.20
04(27)
Europe
F39
–65
181,36
174
Confirm
edcases
Totalcoffee
Age,sm
oking,low
physicalactivity,
education,B
MI,and
serum
cholesteroland
triglycerides
8
#2
1.00(1.00–
1.00
)
3–4
0.56(0.32–
0.98
)
5–6
0.45(0.23–
0.90
)
.6
0.57(0.26–
1.29
)
Salazar-Martinez
etal.20
04(4)
U.S.
Both
M40
–75
M12
41,934
1,33
3Confirm
edCaffeine(m
en)b
Age;totalcaloricintake;
family
history
of
diabetes;alcohol
consumption;sm
oking
status;men
opausal
statusand
postmen
opausal
horm
oneuse;intakes
ofglycem
icload
,trans
fat,polyunsaturated
fattyacid,cerealfi
ber,
andmagnesium;BMI;
andphysicalactivity
6
F30
–55
F18
self-rep
ort
131.00(1.00–
1.00
)
741.06(0.89–
1.26
)
172
1.01
(0.85–1.20)
323
0.94(0.78–
1.12
)
566
0.80(0.66–
0.97
)
M84
,276
Caffeine(w
omen
)b
F4,08
569
1.00(1.00–
1.00
)
193
1.02(0.93–
1.11
)
328
0.90(0.82–
0.99
)
432
0.85(0.77–
0.94
)
708
0.70(0.63–
0.79
)
Con
tinu
edon
p.57
3
572 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
vanDam
etal.20
04(28)
Europe
Both
50–74
6.4
1,31
212
8OGTT
Totalcoffee
Age;sex;B
MI;waist-hip
ratio;p
hysicalactivity;
alcoholconsumption;
cigarettesm
oking;
history
of
cardiovascular
disease;
useof
antihypertensive
medication;
andintake
offiber,totalenergy,
saturatedfat,and
polyunsaturated
fat
8
#2
1.00
(1.00–1.00
)
3–4
0.94
(0.56–1.55
)
5–6
0.92
(0.53–1.61
)
$7
0.69
(0.31–1.51
)
Greenberget
al.20
05(35)
U.S.
Both
32–88
8.4
7,00
617
0Self-rep
orted
Caffeinated
coffee
Percapitaincome,
educationallevel,race,
sex,physicalactivity,
smokingstatus,
alcoholconsumption,
BMI,age,
andtypeof
diet;variablesfor
consumptionof
caffeineandallthree
beverages
werein
the
regressionmodel
simultaneo
usly
6
01.00
(1.00–1.00
)
,2
0.82
(0.55–1.23
)
240.75
(0.50–1.13
)
$4
0.37
(0.22–0.64
)
Decaf.coffee
01.00
(1.00–1.00
)
,2
0.62
(0.34–1.11
)
2–4
0.43
(0.20–0.93
)
Caffeineb
751.00
(1.00–1.00
)
225
0.69
(0.42–1.11
)
375
0.73
(0.38–1.41
)
525
0.65
(0.30–1.42
)
750
0.65
(0.24–1.77
)
Isoet
al.20
06(9)
Asia
Both
40–65
517
,413
444
Self-rep
orted
Totalcoffee
Family
history
ofdiabetes,
smokingstatus,
alcoholintake,
magnesium
intake,
hoursofw
alking,hours
ofexercise,and
consumptionofother
beverages
6
,2
0.82
(0.55–1.23
)
2–4
0.75
(0.50–1.13
)
$4
0.37
(0.22–0.64
)
$3
0.58
(0.37–0.90
)
Caffeineb
571.00
(1.00–1.00
)
137
1.00
(0.73–1.37
)
199
0.98
(0.73–1.33
)
273
1.00
(0.74–1.37
)
416
0.67
(0.47–0.95
)
Con
tinu
edon
p.57
4
care.diabetesjournals.org Ding and Associates 573
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
Paynteret
al.20
06(29)
U.S.
Both
45–64
912
,204
1,43
7Self-rep
orted
Totalcoffee
(men
)Age,race,ed
ucation,
family
history
of
diabetes,BMI,waist-
hip
ratio,totalcaloric
intake,d
ietary
fiber,
smoking,alcohol
consumption,leisure
activity,and
hypertension
8
01.00
(1.00–1.00)
,1
1.02
(0.76–1.38)
11.01
(0.76–1.34)
2–3
0.93
(0.70–1.22)
$4
0.69
(0.50–0.96)
Totalcoffee
(women
)
01.00
(1.00–1.00)
,1
1.23
(0.93–1.62)
11.06
(0.83–1.36)
2–3
0.89
(0.68–1.16)
$4
0.66
(0.45–0.96)
Pereiraet
al.20
06(13)
U.S.
FMean61
1128
,812
1,41
8Self-rep
orted
Totalcoffee
Age,ed
ucation,baseline
hypertension,alcohol,
smokingstatus,
cigarettepack-years,
BMI,waist-hip
ratio,
physicalactivity,
energy
intake,totalfat,
Keys
score,cereal
fiber,tea
consumption,and
sodaconsumption
6
01.00
(1.00–1.00)
,1
0.95
(0.77–1.18)
1–3
1.01
(0.85–1.19)
4–5
0.85
(0.69–1.04)
$6
0.79
(0.61–1.02)
Caffeinated
coffee
01.00
(1.00–1.00)
,1
0.92
(0.76–1.11)
1–3
0.88
(0.70–1.12)
4–5
0.89
(0.64–1.23)
$6
1.00
(0.84–1.19)
Decaf.coffee
01.00
(1.00–1.00)
,1
0.98
(0.83–1.16)
1–3
1.01
(0.84–1.21)
4–5
0.59
(0.44–0.80)
$6
0.68
(0.43–1.09)
Con
tinu
edon
p.57
5
574 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
vanDam
etal.20
06(5)
U.S.
F26
–46
2088
,259
1,263
Self-rep
orted
Totalcoffee
Age;sm
okingstatus;BMI;
physicalactivity;
alcoholconsumption;
use
ofhorm
one
replacementtherapy;
oralcontracep
tive
use;
family
history
oftype2
diabetes;history
of
hypertension
;historyof
hypercholesterolemia;
consumptionofsugar-
sweetened
softdrinks;
consumptionofpun
ch;
andquintilesof
processedmeat
consumption,the
polyunsaturated-to-
saturatedfatintake
ratio,totalenergy
intake,theglycem
ic
index,andcerealfiber
intake
5
01.00
(1.00–1.00)
,1
0.93
(0.80–1.09)
10.87
(0.73–1.03)
2–3
0.58
(0.49–0.68)
$4
0.53
(0.41–0.68)
Caffeinated
coffee
01.00
(1.00–1.00)
,1
1.00
(0.86–1.17)
10.89
(0.75–1.07)
2–3
0.62
(0.52–0.74)
$4
0.61
(0.46–0.81)
Decaf.coffee
01.00
(1.00–1.00)
,1
0.86
(0.74–0.99)
10.87
(0.68–1.11)
2–3
0.52
(0.36–0.74)
Caffeineb
221.00
(1.00–1.00)
930.88
(0.75–1.04)
180
0.89
(0.75–1.05)
341
0.74
(0.62–0.89)
528
0.55
(0.45–0.67)
Smithet
al.20
06(30)
U.S.
Both
$50
891
084
OGTT
Totalcoffee
Sex,age,
physicalactivity,
BMI,sm
oking,alcohol,
hypertension,and
baselinefastingplasm
a
glucose
7
01.00
(1.00–1.00)
1–2
0.66
(0.38–1.14)
3–4
0.53
(0.26–1.08)
$5
0.60
(0.26–1.40)
Aden
eyet
al.20
07(24)
U.S.
FMean31
10mo
1,744
75OGTT
Caffeinated
coffee
Maternalage,
smoking
duringpregn
ancy,
regularalcoholu
se
before
pregnan
cy,
maternalrace,
prepregn
ancy
BMI,
chronichypertension,
andaveragedaily
caloricintake
6
None
1.00
(1.00–1.00)
Moderate
0.58
(0.33–1.02)
High
0.76
(0.40–1.46)
Con
tinu
edon
p.57
6
care.diabetesjournals.org Ding and Associates 575
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
Fuhrm
anet
al.20
09(31)
U.S.
M35
–79
2.6
4,68
551
9OGTT
Totalcoffeec
Age,BMI,sm
okingstatus,
family
history
of
diabetes,ed
ucation,
alcoholintake,index
of
physicalactivity,and
milk
andsugarintake
7
1–2
1.00(1.00–
1.00
)
30.79(0.69–
1.00
)
$4
0.75(0.58–
0.97
)
Ham
eret
al.20
08(32)
Europe
Both
35–55
11.7
5,82
338
7OGTT
Totalcoffee
Five-yearagecategories,
sex,ethnicity,
employm
entgrad
e,
plusBMIcatego
ry,
waist-hip
ratio,
smoking,sex-specific
alcoholintake
tertiles,
physicalactivity
category,fam
ily
history
ofdiabetes,
hypertension,
cholesterol,total
energy
intake,d
iet
pattern,andmutual
adjustmen
tforall
beveragetype
8
01.00(1.00–
1.00
)
#1
0.83(0.60–
1.14
)
2–3
0.85(0.60–
1.20
)
.3
0.80(0.54–
1.18
)
Decaf.coffee
01.00(1.00–
1.00
)
#1
1.13(0.87–
1.47
)
;2–
30.87(0.58–
1.30
)
.3
0.65(0.36–
1.16
)
Odegaard
etal.20
08(33)
Asia
Both
45–74
5.7
36,908
1,88
9Confirm
edself-
report
Totalcoffee
Age
atrecruitmen
t,year
ofinterview,sex,
dialect,ed
ucation,
hypertension,smoking
status,alcohol
consumption,BMI,
physicalactivity,
dietary
variables,and
magnesium
9
,1
1.00(1.00–
1.00
)
10.94(0.81–
1.09
)
2–3
0.83(0.68–
1.01
)
$4
0.70(0.53–
0.93
)
Bidel
etal.20
08(17)
Europe
Both
35–74
5–20
21,826
862
Confirm
edcases
Totalcoffee
Age,BMI,alcohol
consumption,
smoking,physical
activity,andglucose
tolerance
test
7
0–2
1.00(1.00–
1.00
)
3–4
0.85(0.70–
1.04
)
5–6
0.78(0.65–
0.95
)
$7
0.64(0.51–
0.80
)
Con
tinu
edon
p.57
7
576 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
Kato
etal.20
09(34)
Asia
Both
40–69
1055
,826
2,69
4Confirm
ed
self-rep
ort
Totalcoffee
(men
)
Age,BMI,sm
okingstatus,
alcoholdrinking,family
history
ofdiabetes,
physicalactivity,
history
of
hypertension,
perceived
men
tal
stress,levelsoftypeA
beh
avior,andhoursof
sleep
8
01.00
(1.00–1.00)
1–2d
0.93
(0.80–1.08)
3–4d
0.84
(0.71–1.01)
1–2
0.84
(0.73–0.97)
3–4
0.83
(0.68–1.02)
$5
0.82
(0.60–1.11)
Totalcoffee
(women
)
01.00
(1.00–1.00)
1–2d
0.90
(0.76–1.06)
3–4d
0.95
(0.77–1.17)
1–2
0.81
(0.69–0.96)
3–4
0.62
(0.45–0.84)
$5
0.40
(0.20–0.78)
Obaet
al.20
10(10)
Asia
Both
,70
1013
,540
453
Self-rep
orted
Caffeinated
coffee
(men
)
Age,sm
okingstatus,BMI,
physicalactivity,
lengthofed
ucationin
years,alcohol
consumption,total
energy
intake,fat
intake,andwomen
’s
men
opausalstatusto
exam
inebeverageand
chocolate
snacks
6
01.00
(1.00–1.00)
1/moto
6/wk
0.69
(0.50–0.97)
$1
0.69
(0.49–0.98)
Caffeinated
coffee
(women
)
01.00
(1.00–1.00)
1/moto
6/wk
1.08
(0.74–1.60)
$1
0.70
(0.44–1.12)
Decaf.coffee
(men
)
01.00
(1.00–1.00)
$1e
1.09
(0.73–1.61)
Decaf.coffee
(women
)
01.00
(1.00–1.00)
$1e
0.66
(0.36–1.23)
Caffeine(m
en)b
421.00
(1.00–1.00)
790.81
(0.60–1.10)
218
0.95
(0.69–1.30)
Caffeine(w
omen
)b
601.00
(1.00–1.00)
971.26
(0.88–1.82)
226
0.95
(0.63–1.43)
Con
tinu
edon
p.57
8
care.diabetesjournals.org Ding and Associates 577
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffeeexposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
Zhan
get
al.20
11(23)
U.S.
Both
45–74
7.6
1,14
118
8OGTT
Totalcoffee
Age,sex,smoking,alcohol
use,family
history
of
diabetes,physical
activity,andBMI
7
01.00(1.00–
1.00
)
1–2
0.93(0.55–
1.57
)
3–4
0.87(0.53–
1.44
)
5–7
0.72(0.43–
1.23
)
8–11
0.78(0.44–
1.37
)
$12
0.33(0.13–
0.81
)
vanDierenet
al.20
09(22)
Europe
Both
49–70
1040
,011
918
Confirm
ed
self-rep
ort
Totalcoffee
Cohort,sex,age,
BMI,
highested
ucation,
physicalactivity,fam
ily
history
ofdiabetes,
smoking,alcohol
intake,e
nergy
intake,
energy-adjusted
saturatedfatintake,
energy-adjusted
fiber
intake,e
nergy-
adjusted
vitamin
C
intake,
hypercholesterolemia,
andhypertension
6
0–1
1.00(1.00–
1.00
)
1–2
0.88(0.69–
1.11
)
2–3
0.94(0.75–
1.17
)
3–4
0.75(0.60–
0.92
)
4–6
0.74(0.61–
0.91
)
.6
0.84(0.65–
1.08
)
Boggset
al.20
10(21)
U.S.
F30
–69
1246
,906
3,67
1Self-rep
orted
Caffeinated
coffee
Age,questionnaire
cycle,
energy
intake,
education,fam
ily
history
ofdiabetes,
vigo
rousactivity,
smoking,glycem
ic
index,cerealfiber,
sugar-sw
eetened
soft
drinks,B
MI,history
of
hypertension,and
history
ofhigh
cholesterol
6
0e1.00(1.00–
1.00
)
1cup/m
oto
,1cup/day
0.94(0.86–
1.04
)
10.90(0.81–
1.01
)
2–3
0.82(0.72–
0.93
)
$4
0.83(0.69–
1.01
)
Decaf.coffee
0e1.00(1.00–
1.00
)
1cup/m
oto
,1cup/day
1.04(0.93–
1.16
)
11.07(0.92–
1.25
)
2–3
1.01(0.83–
1.24
)
$4
1.10(0.81–
1.49
)
Con
tinu
edon
p.57
9
578 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
Sartorelliet
al.20
10(20)
Europe
F41
–72
1169
,532
1,41
5Confirm
ed
self-rep
ort
Totalcoffee
Age,h
istory
ofdiabetes
in
ascendants,q
uartiles
ofphysicalactivity
level,quartilesof
alcoholintake,
educationallevel,
hypercholesterolemia,
hypertension,
smoking,en
ergy-
adjusted
fiber
and
saturatedfat,total
energy
without
alcohol,men
opausal
status,horm
one
replacemen
ttherapy
asatime-dep
enden
t
variab
le,use
oforal
contracep
tivesat
baseline,
andtime-
dep
enden
tBMI
7
01.00
(1.00–1.00
)
#1
1.04
(0.87–1.26
)
1.1–
2.9
0.86
(0.73–1.02
)
$3
0.73
(0.61–0.87
)
Caffeinated
coffee
01.00
(1.00–1.00
)
#1
0.85
(0.74–0.97
)
.1
0.67
(0.57–0.78
)
Decaf.coffee
01.00
(1.00–1.00
)
#1
0.69
(0.51–0.93
)
.1
0.67
(0.47–0.95
)
Caffeineb
481.00
(1.00–1.00
)
126
0.84
(0.72–0.96
)
209
0.77
(0.66–0.89
)
397
0.67
(0.58–0.78
)
Goto
etal.20
11(12)
U.S.
FMean60
1071
835
9Confirm
ed
self-rep
ort
Caffeinated
coffee
Age,race,d
urationof
follow-up,timeof
blooddraw,smoking
status,physical
activity,family
history
ofdiabetes,alcohol
use,totalcalories,and
BMI
7
01.00
(1.00–1.00
)
#1
0.92
(0.46–1.84
)
2–3
0.96
(0.48–1.94
)
$4
0.71
(0.31–1.61
)
Decaf.coffee
01.00
(1.00–1.00
)
#1
1.31
(0.84–2.03
)
2–3
0.91
(0.42–2.00
)
$4
0.92
(0.26–3.27
)
Caffeineb
131.00
(1.00–1.00
)
140
1.53
(0.71–3.30
)
366
1.70
(0.79–3.62
)
656
1.32
(0.61–2.82
)
Hjellviket
al.20
11(6)
Europe
Both
40–45
5–20
362,045
9,88
6Confirm
edcases
Totalcoffee
Year
ofbirth
and,where
relevant,sex,BMI,
smoking,ed
ucation,
andphysicalactivity
7
,1
1.00
(1.00–1.00
)
1–4
0.84
(0.79–0.89
)
5–8
0.65
(0.61–0.70
)
.9
0.63
(0.58–0.69
)
Con
tinu
edon
p.58
0
care.diabetesjournals.org Ding and Associates 579
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
Floegelet
al.20
12(8)
Europe
Both
35–65
8.9
42,659
1,43
2Confirm
ed
self-rep
ort
Caffeinated
coffee
Age,center,sex,sm
oking,
alcoho
lintake,physical
activity,education,
employment,vitamin
andmineralsupplement
useduring
past4wk,
totalenergyintake,tea
intake,decaf.coffee
intake,BMI,waist-hip
ratio,andprevalent
hypertension
7
,1
1.00
(1.00–1.00
)
1–2
0.89
(0.69–1.16
)
2–3
0.92
(0.76–1.13
)
3–4
0.82
(0.65–1.02
)
$4
0.77
(0.63–0.94
)
Decaf.coffee
,1
1.00
(1.00–1.00
)
1–2
0.97
(0.68–1.39
)
2–3
1.11
(0.84–1.48
)
3–4
0.70
(0.41–1.19
)
$4
0.70
(0.46–1.06
)
Bhupathirajuet
al.20
13(19)
U.S.
Both
M40
–75
F30
–55
M22
F24
39,059
2,86
5Confirm
ed
self-rep
ort
Caffeinated
coffee(m
en)
Age;totalcaloricintake;
family
history
of
diabetes;alcohol
consumption;sm
oking
status;men
opausal
status;
postmen
opausal
horm
oneuse;intake
of
glycem
icload
,trans
fat,polyunsaturated
fattyacid,cerealfiber,
andmagnesium;BMI;
andphysicalactivity
6
,1
1.00
(1.00–1.00
)
1–3
0.95
(0.88–1.03
)
4–5
0.92
(0.82–1.05
)
$6
0.65
(0.49–0.85
)
Caffeinated
coffee
(women
)
,1
1.00
(1.00–1.00
)
1–3
0.91
(0.86–0.95
)
4–5
0.78
(0.72–0.85
)
$6
0.65
(0.58–0.73
)
M74
,749
F7,370
Decaf.coffee
(men
)
01.00
(1.00–1.00
)
,1
0.92
(0.84–1.01
)
1–3
0.84
(0.75–0.93
)
$4
0.84
(0.68–1.03
)
Decaf.coffee
(women
)
01.00
(1.00–1.00
)
,1
0.95
(0.89–1.00
)
1–3
0.86
(0.81–0.92
)
$4
0.73
(0.64–0.84
)
Con
tinu
edon
p.58
1
580 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
Table
1—Continued
Authors
Country
Sex
Age
at
baseline
(year)
Follo
w-up(year)
Nn
Type2
diabetes
assessmen
t
Coffee
exposure
(cups/day)
RR
(95%
CI)
Confoundingadjustmen
ts
NOS
(0–9points)
Dooet
al.20
13(18)
U.S.
Both
45–75
1475
,140
8,58
2Confirm
edcases
Totalcoffee
(men
)Age
andadjusted
for
ethnicity,BMI,physical
activity,ed
ucation,
history
of
hypertension,en
ergy,
alcoholintake,
smokingstatus,
sugaredsodas,d
ietary
fiber
per
4,18
4kJ,and
processed
meatper
4,184kJ
9
01.00
(1.00–1.00
)
,1
1.12
(1.01–1.24
)
11.06
(0.97–1.17
)
21.05
(0.95–1.16
)
$3
0.95
(0.84–1.08
)
Totalcoffee(w
omen
)
01.00
(1.00–1.00
)
,1
1.08
(0.98–1.19
)
10.96
(0.87–1.05
)
20.82
(0.74–0.91
)
$3
0.69
(0.59–0.81
)
Caffeinated
coffee
(men
)
01.00
(1.00–1.00
)
,1
1.07
(0.98–1.18
)
11.02
(0.93–1.11
)
20.99
(0.91–1.09
)
$3
0.86
(0.75–0.98
)
Caffeinated
coffee
(women
)
01.00
(1.00–1.00
)
,1
1.02
(0.94–1.12
)
10.93
(0.86–1.01
)
20.75
(0.68–0.83
)
$3
0.65
(0.54–0.78
)
Decaf.coffee
(men
)
01.00
(1.00–1.00
)
,1
1.08
(1.00–1.16
)
11.03
(0.93–1.15
)
$2
1.07
(0.93–1.23
)
Decaf.coffee
(women
)
01.00
(1.00–1.00
)
,1
1.07
(0.99–1.15
)
11.03
(0.92–1.14
)
$2
0.85
(0.72–1.01
)
Decaf.,decaffeinated
;F,female;
M,m
ale;
mo,month;OGGT,oralglucose
tolerance
test;wk,week.
aIn
person-years.bIn
mg/day.c Inservings/day.dIn
cups/week.
eIn
cups/month.
care.diabetesjournals.org Ding and Associates 581
compared with the lowest category.Compared with the lowest category(median consumption 0 cups/day), thepooled RR for incident type 2 diabetesfor individuals in the highest categoryof consumption (5 cups/day) was 0.70(95% CI 0.65–0.75, I2 = 50%, P forheterogeneity = 0.001). Thecorresponding RRs were 0.80 (0.75–0.85, I2 = 71%, P , 0.001) for thesecond highest category (3.5 cups/day)and 0.91 (0.88–0.94, I2 = 19%, P = 0.17)for the third highest category (1 cup/day) of coffee consumption. Thus,there was evidence for substantialbetween-study heterogeneity inresults for the highest two categoriesof coffee consumption.
Dose-Response Analyses of TotalCoffee Consumption and Risk ofDiabetesWe performed a dose-response meta-analysis in 27 studies to test for alinear trend between total coffeeconsumption and risk of type 2 diabetesand to estimate RRs for specific amountsof coffee consumption. One studyincluded coffee consumption of .12
cups/day for the highest category (23);we treated this observation as an outlierand excluded it from the dose-responseanalysis. There was a strong inverseassociation between coffeeconsumption and risk of type 2 diabetes(Fig. 3). A cubic spline model accountedfor more variance in the outcome thandid a linear model (likelihood ratio testP , 0.001), suggesting that theassociation was not fully linear.Compared with participants with nocoffee consumption, the RR estimateddirectly from the cubic spline modelfor 1–6 cups/day was 0.92 (95% CI0.90–0.94), 0.85 (0.82–0.88), 0.79(0.75–0.83), 0.75 (0.71–0.80), 0.71(0.65–0.76), and 0.67 (0.61–0.74),respectively.
To test whether the inverse associationbetween coffee consumption and riskof type 2 diabetes was different forunadjusted and adjusted RRs, weperformed a dose-response meta-analysis of the unadjusted data in 27studies (Supplementary Fig. 1). Thespline curve of the unadjusted datawas similar to that of the
multivariable-adjusted data, indicatingthat adjustment for potentialconfounders minimally affected effectestimates for the association betweencoffee consumption and a lower risk ofdiabetes.
Comparison of Associations forCaffeinated and Decaffeinated CoffeeBoth caffeinated and decaffeinatedcoffee consumption was inverselyassociated with risk of type 2 diabetes(Supplementary Fig. 2). Compared withparticipants with the lowest level ofcaffeinated coffee consumption, the RRfor incident type 2 diabetes was 0.74(95% CI 0.67–0.81, I2 = 56%, P forheterogeneity = 0.005) for the highestcategory (median consumption 5 cups/day), 0.82 (0.75–0.91, I2 = 74%,P , 0.001) for the second highestcategory (3.5 cups/day), and 0.92 (0.89–0.96, I2 = 6.6%, P = 0.38) for the thirdhighest category (1 cup/day). Comparedwith those with the lowest level ofdecaffeinated coffee consumption, thecombined RR for incident type 2diabetes was 0.80 (0.70–0.91, I2 = 62%,P = 0.001) for the highest category
Figure 2—Forest plot of the associations between total coffee consumption and risk of type 2 diabetes. Compared with the lowest category (medianconsumption 0 cups/day), the pooled RR for incident type 2 diabetes was 0.70 (95% CI 0.65–0.75, I2 = 50%, P for heterogeneity = 0.001) for thehighest category of consumption, 0.80 (0.75–0.85, I2 = 71%, P, 0.001) for the second highest, and 0.91 (0.88–0.94, I2 = 19%, P = 0.17) for the thirdhighest category of coffee consumption.
582 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
(4 cups/day), 0.95 (0.88–1.02, I2 = 53%,P = 0.02) for the second highest category(2 cups/day), and 0.98 (0.92–1.05, I2 =62%, P = 0.003) for the third highestcategory (0.5 cups/day). Heterogeneitywas shown for five of the six subgroups.Seven of 12 studies adjusted forcaffeinated and decaffeinated coffeeconsumption simultaneously. Theassociation was slightly stronger forcaffeinated coffee consumption than fordecaffeinated coffee consumption (P =0.03 for the second highest group, P =0.07 for the highest group)(Supplementary Table 1).
We conducted a linear dose-responseanalysis for caffeinated anddecaffeinated coffee consumptionseparately (11 studies). For a 1 cup/dayincrease in caffeinated coffeeconsumption, the RR for type 2 diabeteswas 0.91 (95% CI 0.89–0.94), and for a1 cup/day increase in decaffeinatedcoffee consumption, the RR was 0.94(0.91–0.98, P for difference = 0.17). Weperformed a dose-response analysisbetween caffeine consumption and type2 diabetes risk based on seven of theincluded studies (4,9,12,17,21,22,35)and found that for every 140 mg/day(;1 cup/day coffee) higher caffeineconsumption, the RR for type 2 diabeteswas 0.92 (0.90–0.94). Of note, none of
the included studies controlled forcoffee intake when modeling caffeineintake and diabetes.
Stratified AnalysisIn stratified analyses, the inverseassociations between coffee consumptionand risk of diabetes were similar bygeographical region (U.S., Europe, andAsia), sex, and diabetes assessmentmethod (P for interaction. 0.05 for allgroups) (Supplementary Table 1).
Sensitivity AnalysisFor the dose-response analysis of totalcoffee consumption with diabetes, afterwe included the observation withextremely high coffee consumption(23), the results did not change.Significant heterogeneity was found forthe highest and second highest totalcoffee consumption, which might beattributable to the heterogeneousamount of coffee consumption in thesecategories. The dose-response analysisaccounted for the heterogeneousconsumption amount of individualstudies and showed an appropriategoodness of fit (P = 0.14).
Heterogeneity was also found inseparate analyses of caffeinated anddecaffeinated coffee consumption. Inaddition to considering differentconsumption amounts, we stratified
the analyses by whether thestudies included caffeinated anddecaffeinated coffee consumptionsimultaneously in the same model. Inthese analyses, the heterogeneity ateach level of caffeinated anddecaffeinated coffee consumptiondecreased significantly, with evenmore inverse results for bothcaffeinated and decaffeinated coffeeconsumption (Supplementary Fig. 3).
Publication BiasThe Egger test provided no evidence ofpublication bias at any levels of totalcoffee, caffeinated coffee, anddecaffeinated coffee consumption(Supplementary Table 2).
CONCLUSIONS
The findings from this systematic reviewand meta-analysis, based on 1,109,272study participants and 45,335 cases oftype 2 diabetes, demonstrate a robustinverse association between coffeeconsumption and risk of diabetes.Compared with no coffee consumption,consumption of 6 cups/day of coffeewas associated with a 33% lower risk oftype 2 diabetes. Caffeinated coffee anddecaffeinated coffee consumption wereboth associated with a lower risk of type2 diabetes. The association between
Figure 3—Dose-response analysis of the association between coffee consumption and risk of type 2 diabetes. For the overall association betweencoffee consumption and risk of diabetes, P, 0.001; for the goodness of fit of themodel, P = 0.14; and for the likelihood ratio test compared with thenested linear model, P , 0.001.
care.diabetesjournals.org Ding and Associates 583
coffee consumption and diabetes riskwas consistent for men and women andfor European, U.S., and Asianpopulations.
There has been some debate about therole of caffeine in the development ofinsulin resistance and diabetes. Short-term studies in humans have shown thatcaffeine intake results in an acutereduction of insulin sensitivity, whichmay be due to adenosine receptorantagonism and increased epinephrinerelease caused by caffeine (36). Otherstudies have found that whencontrolling for total coffee intake,caffeine intake was no longer associatedwith diabetes risk (5,19). Nonetheless,these results need to be interpretedwith caution because the highcorrelation between coffee and caffeineconsumption makes it more difficult toseparate their effects. The presentmeta-analysis shows that bothcaffeinated and decaffeinated coffeeconsumption is associated with a lowerrisk of diabetes. The results were moreinverse when only including the studiesadjusting for caffeinated anddecaffeinated coffee simultaneously inthe model. Because caffeinated coffeeconsumption is inversely related todecaffeinated coffee consumption,caffeinated coffee is a positiveconfounder for the association ofdecaffeinated coffee with diabetes andvice versa. Thus, the association isstrengthened when both variables areadjusted simultaneously. Moreover,because the consumption of caffeinatedcoffee is much higher than decaffeinatedcoffee, the results appear to be morerobust for caffeinated coffee. However,the difference for continuous analysisbetween caffeinated and decaffeinatedcoffee (1 cup/day increment) was notstatistically significant. These resultssuggest that components of coffee otherthan caffeine are responsible for thisputative beneficial effect. Although thepresent results show that higher caffeineconsumption is inversely associated withthe risk of diabetes, this association islikely to be confounded by othercomponents of coffee because of thecollinearity of caffeine and thosecomponents.
Plausible biological mechanisms thatmay contribute to the inverse
association between coffeeconsumption and diabetes risk havebeen demonstrated in animal modelsand in vitro studies implicating severalcoffee components in reducing insulinresistance and improving glucosemetabolism. Chlorogenic acid, aphenolic compound, is a majorcomponent of coffee and has beenshown to reduce blood glucoseconcentrations in animal experiments(37). Chlorogenic acid may reduceglucose absorption in the intestines bycompetitively inhibiting glucose-6-phosphate translocase and reducingsodium-dependent glucose transport inthe brush border membrane vesicles(38), by reducing oxidative stress as aresult of its antioxidant properties, andby reducing liver glucose output (39).Other components of coffee may alsoimprove glucose metabolism, includinglignans, quinides, and trigonelline(40–42). Coffee is also a good source ofmagnesium, which has been associatedwith a lower risk of type 2 diabetes (43).However, adjustment for magnesiumintake did not substantially attenuatethe inverse association (4,5), suggestingthat other bioactive compounds incoffee may also be responsible. Inaddition, because some of thecompounds, such as lignans andmagnesium, also exist in other foods, itis difficult to attribute the benefits ofcoffee on diabetes to these specificcompounds only.
This meta-analysis has severalsignificant strengths. The mean NOSscore of 7 of 9 ensured the relativelyhigh quality of the included studies. All28 studies used a prospective design;thus, the differential misclassificationof coffee consumption attributable torecall bias was minimized. Given theobservational nature of the studies,residual confounding might be anexplanation for the findings. However,higher coffee consumption wasgenerally associated with a less healthylifestyle, including a high prevalence ofcigarette smoking, less physicalactivity, and a less healthy diet. Thus,the true association between coffeeand diabetes risk might be strongerthan observed. Indeed, the similarityfound in the present adjusted andunadjusted results indicates that the
association is unlikely to besubstantially confounded by knowndiabetes risk factors. Nonetheless, it isdifficult to establish the causalitybetween coffee consumption anddiabetes solely based on observationalevidence.
Because most of the studies coffee wasassessed through self-reported dietaryquestionnaires, misclassification ofcoffee intake is inevitable. However, themisclassification is likely to benondifferential with respect to theoutcome because of the prospectivedesign of the studies, which might haveattenuated the observed association.Only five studies addressed the loss tofollow-up percentage. However, the lossto follow-up is unlikely to be associatedwith coffee consumption and may haveresulted in selection bias. Coffeebrewing methods were not assessed inthe studies. However, most coffee islikely to be filtered coffee, and theresults from studies conducted invarious populations, including U.S.,European, and Asian, were similar,indicating consistency of the resultsdespite potentially differentpreparation and processing methods. Inaddition, none of the studies assessedthe amount of sugar and dairy added tocoffee. However, the amount of sugarand dairy added to coffee is likely to besmall compared with other foodsources. Finally, because of theobservational nature of the studies, acausal relationship cannot beestablished with these data alone.
In conclusion, this meta-analysisprovides strong evidence that higherconsumption of coffee is associatedwith a significantly lower risk ofdiabetes. Both caffeinated coffee anddecaffeinated coffee are associatedwith reduced diabetes risk. Longer-termrandomized controlled trials are neededto establish causality and to elucidatethe underlying mechanisms.
Funding. This research was supported byNational Institutes of Health grant DK-58845.
Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.
Author Contributions. M.D. contributed tothe data research, extraction, and analyses andwrote themanuscript.M.D., S.N.B.,M.C., R.M.v.D.,and F.B.H. contributed substantially to the study
584 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014
conception and design, data analysis andinterpretation, and drafting and critical revision ofthe manuscript for important intellectual content.S.N.B. contributed to the manuscript organizationand reviewed and edited the manuscript. M.C.double-checked the extracted data and reviewedthemanuscript. R.M.v.D. and F.B.H. contributed tothe introduction and discussion and reviewed andedited the manuscript. F.B.H. is the guarantor ofthis work and, as such, had full access to all thedata in the study and takes responsibility for theintegrity of the data and the accuracy of the dataanalysis.
References1. Wild S, Roglic G, Green A, Sicree R, King H.
Global prevalence of diabetes: estimatesfor the year 2000 and projections for 2030.Diabetes Care 2004;27:1047–1053
2. Nathan DM. Long-term complications ofdiabetes mellitus. N Engl J Med 1993;328:1676–1685
3. van Dam RM, Feskens EJM. Coffeeconsumption and risk of type 2 diabetesmellitus. Lancet 2002;360:1477–1478
4. Salazar-Martinez E, Willett WC, Ascherio A,et al. Coffee consumption and risk for type2 diabetes mellitus. Ann Intern Med 2004;140:1–8
5. van DamRM,WillettWC,Manson JE, Hu FB.Coffee, caffeine, and risk of type 2diabetes: a prospective cohort study inyounger and middle-aged U.S. women.Diabetes Care 2006;29:398–403
6. Hjellvik V, Tverdal A, Strøm H. Boiled coffeeintake and subsequent risk for type 2diabetes. Epidemiology 2011;22:418–421
7. Carlsson S, Hammar N, Grill V, Kaprio J.Coffee consumption and risk of type 2diabetes in Finnish twins. Int J Epidemiol2004;33:616–617
8. Floegel A, Pischon T, Bergmann MM,Teucher B, Kaaks R, Boeing H. Coffeeconsumption and risk of chronic diseasein the European Prospective Investigationinto Cancer and Nutrition (EPIC)-Germany study. Am J Clin Nutr 2012;95:901–908
9. Iso H, Date C, Wakai K, Fukui M, TamakoshiA; JACC Study Group. The relationshipbetween green tea and total caffeine intakeand risk for self-reported type 2 diabetesamong Japanese adults. Ann Intern Med2006;144:554–562
10. Oba S, Nagata C, Nakamura K, et al.Consumption of coffee, green tea, oolongtea, black tea, chocolate snacks and thecaffeine content in relation to risk ofdiabetes in Japanese men and women. Br JNutr 2010;103:453–459
11. Reunanen A, Heliovaara M, Aho K. Coffeeconsumption and risk of type 2 diabetesmellitus. Lancet 2003;361:702–703; authorreply 703
12. Goto A, Song Y, Chen BH,Manson JE, BuringJE, Liu S. Coffee and caffeine consumption
in relation to sex hormone-binding globulinand risk of type 2 diabetes inpostmenopausal women. Diabetes 2011;60:269–275
13. Pereira MA, Parker ED, Folsom AR. Coffeeconsumption and risk of type 2 diabetesmellitus: an 11-year prospective study of 28812 postmenopausal women. Arch InternMed 2006;166:1311–1316
14. van Dam RM, Hu FB. Coffee consumptionand risk of type 2 diabetes: a systematicreview. JAMA 2005;294:97–104
15. Huxley R, Lee CM, Barzi F, et al. Coffee,decaffeinated coffee, and tea consumptionin relation to incident type 2 diabetesmellitus: a systematic review with meta-analysis. Arch Intern Med 2009;169:2053–2063
16. Hu G, Jousilahti P, Peltonen M, Bidel S,Tuomilehto J. Joint association of coffeeconsumption and other factors to the riskof type 2 diabetes: a prospective study inFinland. Int J Obes (Lond) 2006;30:1742–1749
17. Bidel S, Silventoinen K, Hu G, Lee DH, KaprioJ, Tuomilehto J. Coffee consumption, serumgamma-glutamyltransferase and risk oftype II diabetes. Eur J Clin Nutr 2008;62:178–185
18. Doo T, Morimoto Y, Steinbrecher A,Kolonel LN, Maskarinec G. Coffee intakeand risk of type 2 diabetes: theMultiethnic Cohort. Public Health Nutr. 27February 2013 [Epub ahead of print]
19. Bhupathiraju SN, Pan A, Malik VS, et al.Caffeinated and caffeine-free beveragesand risk of type 2 diabetes. Am J Clin Nutr2013;97:155–166
20. Sartorelli DS, Fagherazzi G, Balkau B, et al.Differential effects of coffee on the risk oftype 2 diabetes according to mealconsumption in a French cohort of women:the E3N/EPIC cohort study. Am J Clin Nutr2010;91:1002–1012
21. Boggs DA, Rosenberg L, Ruiz-Narvaez EA,Palmer JR. Coffee, tea, and alcohol intake inrelation to risk of type 2 diabetes in AfricanAmerican women. Am J Clin Nutr 2010;92:960–966
22. van Dieren S, Uiterwaal CS, van der SchouwYT, et al. Coffee and tea consumption andrisk of type 2 diabetes. Diabetologia 2009;52:2561–2569
23. Zhang Y, Lee ET, Cowan LD, Fabsitz RR,Howard BV. Coffee consumption and theincidence of type 2 diabetes in men andwomen with normal glucose tolerance: theStrong Heart Study. Nutr Metab CardiovascDis 2011;21:418–423
24. Adeney KL, Williams MA, Schiff MA, Qiu C,Sorensen TK. Coffee consumption and therisk of gestational diabetes mellitus. ActaObstet Gynecol Scand 2007;86:161–166
25. Higgins JP, Thompson SG, Deeks JJ,Altman DG. Measuring inconsistency
in meta-analyses. BMJ 2003;327:557–560
26. Greenland S, Longnecker MP. Methodsfor trend estimation from summarizeddose-response data, with applications tometa-analysis. Am J Epidemiol 1992;135:1301–1309
27. Rosengren A, Dotevall A, Wilhelmsen L,Thelle D, Johansson S. Coffee and incidenceof diabetes in Swedish women:a prospective 18-year follow-up study.J Intern Med 2004;255:89–95
28. van Dam RM, Dekker JM, Nijpels G,Stehouwer CD, Bouter LM, Heine RJ;Hoorn study. Coffee consumption andincidence of impaired fasting glucose,impaired glucose tolerance, and type 2diabetes: the Hoorn Study. Diabetologia2004;47:2152–2159
29. Paynter NP, Yeh HC, Voutilainen S, et al.Coffee and sweetened beverageconsumption and the risk of type 2 diabetesmellitus: the atherosclerosis risk incommunities study. Am J Epidemiol 2006;164:1075–1084
30. Smith B, Wingard DL, Smith TC, Kritz-Silverstein D, Barrett-Connor E. Does coffeeconsumption reduce the risk of type 2diabetes in individuals with impairedglucose? Diabetes Care 2006;29:2385–2390
31. Fuhrman BJ, Smit E, Crespo CJ, Garcia-Palmieri MR. Coffee intake and risk ofincident diabetes in Puerto Rican men:results from the Puerto Rico Heart HealthProgram. Public Health Nutr 2009;12:842–848
32. Hamer M, Witte DR, Mosdøl A, MarmotMG, Brunner EJ. Prospective study of coffeeand tea consumption in relation to risk oftype 2 diabetes mellitus among men andwomen: the Whitehall II study. Br J Nutr2008;100:1046–1053
33. Odegaard AO, Pereira MA, Koh WP,Arakawa K, Lee HP, Yu MC. Coffee, tea, andincident type 2 diabetes: the SingaporeChinese Health Study. Am J Clin Nutr 2008;88:979–985
34. Kato M, Noda M, Inoue M, Kadowaki T,Tsugane S; JPHC Study Group. Psychologicalfactors, coffee and risk of diabetes mellitusamong middle-aged Japanese:a population-based prospective study inthe JPHC study cohort. Endocr J 2009;56:459–468
35. Greenberg JA, Axen KV, Schnoll R, BoozerCN. Coffee, tea and diabetes: the role ofweight loss and caffeine. Int J Obes (Lond)2005;29:1121–1129
36. Thong FS, Graham TE. Caffeine-inducedimpairment of glucose tolerance isabolished by beta-adrenergic receptorblockade in humans. J Appl Physiol (1985)2002;92:2347–2352
37. van Dam RM. Coffee consumption and riskof type 2 diabetes, cardiovascular diseases,
care.diabetesjournals.org Ding and Associates 585
and cancer. Appl Physiol Nutr Metab 2008;33:1269–1283
38. Arion WJ, Canfield WK, Ramos FC, et al.Chlorogenic acid andhydroxynitrobenzaldehyde: new inhibitorsof hepatic glucose 6-phosphatase. ArchBiochem Biophys 1997;339:315–322
39. Svilaas A, Sakhi AK, Andersen LF, et al. Intakesof antioxidants in coffee,wine, and vegetables
are correlated with plasma carotenoids inhumans. J Nutr 2004;134:562–567
40. van Dam RM. Coffee and type 2 diabetes:from beans to beta-cells. Nutr MetabCardiovasc Dis 2006;16:69–77
41. van Dijk AE, Olthof MR, Meeuse JC, SeebusE, Heine RJ, van Dam RM. Acute effects ofdecaffeinated coffee and the major coffeecomponents chlorogenic acid and
trigonelline on glucose tolerance. DiabetesCare 2009;32:1023–1025
42. Greenberg JA, Boozer CN, Geliebter A.Coffee, diabetes, and weight control. Am JClin Nutr 2006;84:682–693
43. Lopez-Ridaura R, Willett WC, Rimm EB,et al. Magnesium intake and risk of type 2diabetes in men and women. Diabetes Care2004;27:134–140
586 Coffee and Risk of Type 2 Diabetes Diabetes Care Volume 37, February 2014