ARTICLE

Consumption of sweet beverages and type 2 diabetesincidence in European adults: results from EPIC-InterAct

The InterAct consortium

Received: 21 December 2012 /Accepted: 11 March 2013 /Published online: 26 April 2013# Springer-Verlag Berlin Heidelberg 2013

AbstractAims/hypothesis Consumption of sugar-sweetened beverageshas been shown, largely in American populations, to increasetype 2 diabetes incidence.We aimed to evaluate the associationof consumption of sweet beverages (juices and nectars, sugar-sweetened soft drinks and artificially sweetened soft drinks)with type 2 diabetes incidence in European adults.Methods We established a casecohort study including 12,403incident type 2 diabetes cases and a stratified subcohort of16,154 participants selected from eight European cohortsparticipating in the European Prospective Investigationinto Cancer and Nutrition (EPIC) study. After exclusions, thefinal sample size included 11,684 incident cases and asubcohort of 15,374 participants. Cox proportional hazardsregression models (modified for the casecohort design)and random-effects meta-analyses were used to estimate theassociation between sweet beverage consumption (obtainedfrom validated dietary questionnaires) and type 2 diabetesincidence.Results In adjusted models, one 336 g (12 oz) daily incrementin sugar-sweetened and artificially sweetened soft drinkconsumption was associated with HRs for type 2 diabetes of1.22 (95% CI 1.09, 1.38) and 1.52 (95% CI 1.26, 1.83),respectively. After further adjustment for energy intake andBMI, the association of sugar-sweetened soft drinks with type2 diabetes persisted (HR 1.18, 95% CI 1.06, 1.32), but the

association of artificially sweetened soft drinks becamestatistically not significant (HR 1.11, 95% CI 0.95, 1.31).Juice and nectar consumption was not associated with type 2diabetes incidence.Conclusions/interpretation This study corroborates theassociation between increased incidence of type 2 diabetesand high consumption of sugar-sweetened soft drinks inEuropean adults.

Keywords Artificially sweetened soft drinks . Casecohortstudy . Europe . Juices and nectars . Prospective study .

Sugar-sweetened soft drinks . Sweet beverages . Type 2diabetes incidence

AbbreviationsEPIC European Prospective Investigation into Cancer

and Nutrition studyMDP Mediterranean dietary patternrMED Relative Mediterranean Diet Score

Introduction

Use of caloric sweeteners (i.e. products used for sweeteningthat are either derived from sugar crops, cereals, fruits ormilk) has increased around the world, with sugar-sweetenedbeverages accounting for a major element of this increase[1]. Parallel to this increase, there has been a rise in theprevalence of obesity and type 2 diabetes worldwide [2].Sugar-sweetened beverage consumption may lead to type 2diabetes because of its effect on weight gain [3], as well asits glycaemic effectsinducing rapid spikes in glucose andinsulin and causing insulin resistance [4, 5]. A meta-analysispublished in 2010 provides empirical evidence for a linkbetween the consumption of sugar-sweetened beverages(defined as drinks containing energy sweeteners, such as

The InterAct Consortium authors are listed in the Appendix.

Electronic supplementary material The online version of this article(doi:10.1007/s00125-013-2899-8) contains peer-reviewed but uneditedsupplementary material, which is available to authorised users.

The InterAct consortium (*)c/o D. Romaguera, Department of Epidemiology & Biostatistics,School of Public Health, Imperial College London,St Marys Campus, Norfolk Place,London W2 1PG, UKe-mail: d.romaguera-bosch@imperial.ac.uk

Diabetologia (2013) 56:15201530DOI 10.1007/s00125-013-2899-8

http://dx.doi.org/10.1007/s00125-013-2899-8

sucrose, high-fructose corn syrup and fruit juice concentrate[i.e. soft drinks and juices with added sugars]) and type 2diabetes [6]. The association between the consumption ofother beverage types (i.e. 100% juices and artificiallysweetened drinks) and type 2 diabetes risk is less clear, withsome studies showing inconsistent results [712].

The majority of epidemiological studies of the associationbetween sweet beverages and type 2 diabetes risk published sofar are based on North American cohorts. The consumption ofsweet beverages in Europe appears to be lower than inthe USA, though it seems to be rising [13]. Data from tenEuropean countries collected in the European ProspectiveInvestigation into Cancer and Nutrition (EPIC) study showedthat, between 1992 and 2000, on average 2.5% of totalcarbohydrate intake came from soft drink consumption [14];the corresponding figure observed in the US Department ofAgriculture survey of 1994/96 was 10.4% [1]. In addition, thediversity of dietary patterns in Europe leads to a wide variationin sweet beverage consumption across different countries,with a clear northsouth gradient (i.e. higher consumption innorthern than in southern European countries) [13, 15].

The aim of the present analysis was to investigate theassociation between the consumption of sweet beverages(i.e. juices and nectars, sugar-sweetened soft drinks andartificially sweetened soft drinks) and the incidence of type2 diabetes in a European cohort, the EPIC-InterAct study.

Methods

Study design The participants, methods, study design andmeasurements have been described previously [16]. In brief,the InterAct project was initiated to investigate how geneticand lifestyle behavioural factors, particularly diet and physicalactivity, interact for the risk of developing diabetes and howknowledge about such interactions may be translated intopreventive action. As part of the wider InterAct project,consortium partners established a casecohort study ofincident type 2 diabetes (InterAct study) based on casesoccurring within 3.99 million person-years accrued between1991 and 2007 in 340,234 people from eight countriesparticipating in EPIC cohorts. All participants gave writteninformed consent, and the study was approved by the localethics committees in the participating countries and theInternal Review Board of the International Agency forResearch on Cancer.

Case ascertainment We followed a pragmatic high-sensitivity approach for case ascertainment with the aimof: (1) identifying all potential incident diabetes cases; and(2) excluding all individuals with prevalent diabetes.

Prevalent diabetes was identified on the basis of baselineself-report of a history of diabetes, doctor-diagnosed

diabetes, diabetes drug use or evidence of diabetes afterbaseline with a date of diagnosis earlier than the baselinerecruitment date. All participants with any evidence ofdiabetes at baseline were excluded.

Ascertainment of incident type 2 diabetes involved areview of the existing EPIC datasets at each centre usingmultiple sources of evidence, including self-report, linkageto primary-care registers, secondary-care registers, medicationuse (drug registers), hospital admissions and mortality data.Information from any follow-up visit or external evidencewith a date later than the baseline visit was used. Cases inDenmark and Sweden were not ascertained by self-report, butidentified via local and national diabetes and pharmaceuticalregisters and hence all ascertained cases were considered to beverified. To increase the specificity of the case definition forcentres other than those from Denmark and Sweden, wesought further evidence for all cases with information onincident type 2 diabetes from fewer than two independentsources at a minimum, including individual medical recordsreviews in some centres. Follow-up was censored at the dateof diagnosis, 31 December 2007, or the date of death,whichever occurred first. In total, 12,403 verified incidentcases were identified during follow-up.

Subcohort The casecohort design of the InterAct studydiffers from the nested casecontrol design in that a randomsubcohort is selected instead of a set of matched controls. Asubcohort of 16,835 individuals were randomly selectedfrom those with available stored blood and buffy coat,stratified by centre. We oversampled the number of individualsin the subcohort for the proportion of prevalent diabetes casesin each centre to account for later exclusion of individuals withprevalent diabetes from InterAct analyses. After exclusion of548 individuals with prevalent diabetes, 129 individualswithout information on reported diabetes status and fourindividuals with post-censoring diabetes, 16,154 subcohortindividuals were included in the analysis. Because ofrandom selection at baseline, this subcohort also included 778individuals who developed incident type 2 diabetes duringfollow-up.

For the present analysis we excluded participants (bothcases and participants in the subcohort) within the lowestand highest 1% of the cohort distribution of the ratio ofreported total energy intake:energy requirement (n=619)and those with missing information on diet (n=117), physicalactivity (n=289), level of education (n=127), smoking status(n=132) or BMI (n=167). This resulted in a final sample sizeof 11,684 type 2 diabetes cases and a subcohort of 15,374(including 730 of the diabetes cases).

Dietary assessment Usual food intake over the previousyear (g/day) was estimated using country-specific validateddietary questionnaires [17] that were administered once at

Diabetologia (2013) 56:15201530 1521

baseline. Estimated individual energy intake (MJ/day) wasderived from foods included in the dietary questionnairesthrough the standardised EPIC Nutrient Database [18].

Sweet beverages evaluated in the present study werejuices and nectars and total soft drinks.

Juices and nectars combines the information collectedon consumption of juices (obtained from either 100% fruitand vegetables or concentrates) and nectars (juices with upto 20% added sugar) across all EPIC countries/centres. Thedistinction of juices from nectars or of 100% fruit or vegetablejuices from juices from concentrates could not be establishedbecause of a lack of standardised information across thedifferent EPIC centres. Therefore, these food items werealways studied in combination.

Total soft drinks combines the information collected onconsumption of carbonate/soft/isotonic drinks and dilutedsyrups across all EPIC countries/centres. The different typesof soft drinks could be distinguished in all EPIC centresexcept Italy, Spain and Ume (in Sweden) and, hence,the variable of total soft drinks could be divided into:(1) sugar-sweetened soft drinks; and (2) artificially sweetenedsoft drinks. Italy, Spain and Ume were excluded from analysesof the association between sugar-sweetened and artificiallysweetened soft drinks and type 2 diabetes incidence.

Sweet beverages were divided into the following categoriesof average consumption: 16 glasses/week; 1 glass/day, with one glass equivalent to250 g (~8.8 oz), the standard serving size used in the EPICdietary questionnaires.

Assessment of other covariates Standard questionnaireswere used to collect information on the participantssociodemographic characteristics and lifestyle variables[19]. For the present analysis we used information about:smoking status (never smoker, former smoker and currentsmoker); alcohol intake (non-drinker, 0.14.9, 4.915, 1530, 3060 and >60 g/day); educational level (no formaleducation, primary school, technical school, secondary schooland university degree); and an ordered four-category index ofphysical activity (inactive, moderately inactive, moderatelyactive and active) [20].

Body weight (kg) and height (cm) were measuredaccording to standardised procedures without shoes,except for the centres at Oxford (UK) and France, whereself-reported anthropometric values were used. Waistcircumference was measured in cm at the narrowest torsocircumference or at the midpoint between the lower ribsand iliac crest in all centres but Ume (Sweden). Weightand waist measurements were corrected to account forprotocol differences between centres in clothing wornduring measurement, as previously described [21]. BMIwas calculated as body weight (kg) divided by heightsquared (m2). Individuals were categorised into groups

of normal weight (BMI

(continuous), presence of hyperlipidaemia (yes, no) andhypercholesterolaemia (yes, no) were performed.

To deal with plausible residual confounding associatedwith the dietary pattern of those with high consumption ofsweet beverages, models were also adjusted for theconsumption of vegetables, fruits, nuts, cereals and products,dairy products, red meat, processed meat, sugar andconfectionary, cakes and biscuits, and coffee and tea (includedin the model as continuous variables expressed in g/day).Also, given that a Mediterranean dietary pattern (MDP) hasbeen shown to be associated with both type 2 diabetesincidence [23] and consumption of sweet beverages [24], weadjusted for the relative Mediterranean Diet Score (rMED;score range, 018), which assesses adherence to MDP,based on reported consumption of nine dietary componentscharacteristic of an MDP: vegetables; legumes; fruits andnuts; cereals; fish and seafood; olive oil; moderate alcoholconsumption; meat and meat products; and dairy products.More information on the construction of rMED can be foundelsewhere [23]. Finally, to minimise reverse causality causedby people who may have changed their dietary habits becauseof an impaired glucose tolerance or chronic disease, weexcluded participants with cardiovascular diseases at baseline(stroke, angina and heart disease), those with family history oftype 2 diabetes, and participants in the first 2 and 5 years offollow-up.

Effect modification by age group (

12 oz increments in total soft drink consumption was1.32 (95% CI 1.20, 1.46). The HR was attenuated to 1.14(95% CI 1.04, 1.26) after further adjustment for BMI. HRswere virtually unchanged in further sensitivity analyses. Astatistically significant interaction (p=0.03) between sex andsoft drink consumption was observed, reflecting a higherHR for women than men; however, the p for interaction wasno longer significant in BMI-adjusted models (p=0.86; notshown in tables). HRs were also higher in younger (

artificially sweetened soft drinks was 1.52 (95% CI 1.26,1.83), which was attenuated and became statistically not sig-nificant after further adjustment for body adiposity measure-ments. A statistically significant interaction (p=0.03) betweenthe consumption of artificially sweetened soft drinks and BMIcategory on type 2 diabetes incidence was observed. In strat-ified analyses, artificially sweetened soft drink consumptionwas significantly associated with diabetes incidence in thenormal-weight participants, and of borderline significance inthe overweight; no association was detected among the obese.The observed significant association in the group with normalweight remained unchanged after further adjustment for BMI(HR 1.43, 95% CI 1.05, 1.95; not shown in tables).

ESM Figs 14 show the HR (95% CI) of type 2 diabetesassociated with 12 oz increments of sweet beverage consump-tion by country/centre. There was evidence of heterogeneity

by country/centre in analyses of total soft drink (I2=63.3%;p=0.002) and artificially sweetened soft drink (I2=60.3%;p=0.01) consumption and diabetes incidence, and moderateheterogeneity with sugar-sweetened soft drink consumption(I2=48.1%; p=0.05). However, the direction of the observedassociations was similar across country/centres.

Discussion

In European men and women, one 12 oz daily increment insugar-sweetened soft drink consumption was associated witha 22% increase in HR for type 2 diabetes, and one 12 oz dailyincrement in artificially sweetened soft drinks was associatedwith a 52% increase in HR. After further adjustment forBMI, the association between sugar-sweetened soft drink

Table 2 HRs (and 95% CIs) for type 2 diabetes according to type and amount of sweet beverage consumption in the EPIC-InterAct study

Variable and model 16 glassesa/week

1 glassa/day

p for trend

HR HRb (95% CI) HRb (95% CI) HRb (95% CI)

Juices and nectars (median intake, g/day) (0.0) (17.1) (100.0) (338.3)

No. cases 5,837 1,702 3,425 720

Crude model 1.00 (ref) 0.88 (0.80, 0.98) 0.89 (0.83, 0.94) 0.97 (0.85, 1.11) 0.64

Adjusted model 1.00 (ref) 0.91 (0.80, 1.02) 0.96 (0.88, 1.04) 1.00 (0.87, 1.15) 0.63

Adjusted model+EI 1.00 (ref) 0.91 (0.81, 1.02) 0.96 (0.88, 1.04) 0.99 (0.86, 1.14) 0.84

Adjusted model+EI+BMI 1.00 (ref) 0.97 (0.86, 1.10) 1.04 (0.96, 1.13) 1.06 (0.90, 1.25) 0.21

Total soft drinksc (median intake, g/day) (0.0) (20.0) (95.1) (413.1)No. cases 5,794 1,604 2,987 1,299

Crude model 1.00 (ref) 1.21 (1.07, 1.36) 1.30 (1.18, 1.43) 1.78 (1.55, 2.04)

Tab

le3

HRs(and

95%

CIs)fortype

2diabetes

associated

with

one33

6g(12oz)increm

entintheconsum

ptionof

differenttyp

eof

sweetb

everages

intheEPIC-InterActstud

y,effect-m

odification

analyses

andsensitivity

analyses

Per

one33

6gservingsize

increm

ent

No.

ofcases

Juices

andnectars

Total

softdrinks

aNo.

ofcasesc

Sug

ar-sweetenedsoftdrinks

Artificially

sweetenedsoftdrinks

HRb(95%

CI)

HRb(95%

CI)

HRb(95%

CI)

HRb(95%

CI)

Unadjustedmod

el11,684

1.00

(0.92,

1.10

)1.43

(1.29,

1.58

)7,116

1.39

(1.21,

1.60

)1.60

(1.32,

1.95

)

Adjustedmod

el11,684

1.04

(0.96,

1.14

)1.32

(1.20,

1.46

)7,116

1.22

(1.09,

1.38

)1.52

(1.26,

1.83

)

Sensitiv

ityanalyses

+Energyintake

11,684

1.02

(0.94,

1.12

)1.32

(1.19,

1.47

)7,116

1.23

(1.08,

1.39

)1.51

(1.25,

1.82

)

+BMI

11,684

1.06

(0.95,

1.18

)1.14

(1.04,

1.26

)7,116

1.18

(1.07,

1.30

)1.12

(0.95,

1.31

)

+Energyintake

+BMI

11,684

1.05

(0.94,

1.18

)1.15

(1.05,

1.27

)7,116

1.18

(1.06,

1.32

)1.11

(0.95,

1.31

)

+Waistcircum

ferenced

10,880

1.04

(0.94,

1.16

)1.17

(1.06,

1.29

)7,116

1.18

(1.05,

1.32

)1.08

(0.97,

1.20

)

+Hyp

ertension

11,684

1.02

(0.93,

1.12

)1.28

(1.16,

1.41

)7,116

1.19

(1.06,

1.33

)1.44

(1.18,

1.76

)

+Hyp

erlip

idaemiae

9,17

91.07

(0.96,

1.18

)1.36

(1.20,

1.55

)5,41

51.30

(1.09,

1.50

)1.45

(1.17,

1.79

)

+Foo

dgrou

psf

11,684

1.05

(0.96,

1.15

)1.29

(1.16,

1.43

)7,116

1.18

(1.05,

1.32

)1.48

(1.20,

1.82

)

+rM

EDg

11,684

1.05

(0.96,

1.15

)1.31

(1.18,

1.45

)7,116

1.20

(1.06,

1.35

)1.53

(1.25,

1.87

)

Excluding

baselin

eCVDh

6,97

71.00

(0.88,

1.13

)1.37

(1.16,

1.63

)3,21

31.37

(1.03,

1.81

)1.36

(1.06,

1.74

)

Excluding

first2yearsof

follo

w-up

10,696

1.01

(0.92,

1.11)

1.32

(1.20,

1.46

)6,12

81.23

(1.09,

1.34

)1.50

(1.24,

1.80

)

Excluding

first5yearsof

follo

w-up

8,21

21.01

(0.89,

1.14

)1.37

(1.23,

1.52

)4,60

91.27

(1.10,

1.46

)1.53

(1.28,

1.82

)

Excluding

family

historyof

diabetes

i3,72

11.09

(0.93,

1.26

)1.30

(1.12,

1.52

)2,91

71.27

(1.03,

1.56

)1.46

(1.15,

1.85

)

Effectmod

ification

Age

grou

p