Diet Drink Consumption and the Risk of Cardiovascular Events:
A Report from the Women’s Health Initiative
Ankur Vyas, MD
, Linda Rubenstein, PhD
, Jennifer Robinson, MD, MPH
Rebecca A. Seguin, PhD, CSCS
, Mara Z. Vitolins, DrPH, MPH, RD
Rasa Kazlauskaite, MD, MSc, FACE
, James M. Shikany, DrPH
, Karen C. Johnson, MD, MPH
Linda Snetselaar, RD, PhD
, and Robert Wallace, MD, MSc
Division of Cardiovascular Medicine, University of Iowa Hospitals & Clinics, Iowa City, IA, USA;
Department of Epidemiology, College of Public Health,
University of Iowa, Iowa City, IA, USA;
Division of Nutritional Sciences, Cornell University, Ithaca, NY, USA;
Department of Epidemiology & Prevention, Wake
Forest School of Medicine, Winston-Salem, NC, USA;
Department of Preventive Medicine, Rush University, Chicago, IL, USA;
Department of Internal
Medicine, Rush University, Chicago, IL, USA;
Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, AL, USA;
Preventive Medicine, University of Tennessee Health Science Center, Memphis, TN, USA;
Department of Internal Medicine, University of Iowa Hospitals and
Clinics, Iowa City, IA, USA.
BACKGROUND: Data are limited regarding the influence
of diet drink consumption on cardiovascular disease
OBJECTIVE: We aimed to evaluate the relationship be-
tween diet drink intake and cardiovascular events.
DESIGN: We conducted a retrospective cohort study, uti-
lizing data from the national, multicenter Women’s Health
Initiative Observational Study (WHI OS), recruiting sub-
jects from 1993 to 1998.
PATIEN T S: Post-menopausal women with available diet
drink intake data, without pre-existing CVD and who
survived ≥60 days were included in the study.
MAIN MEAURES: A composite of incident coronary heart
disease, heart failure, myocardial infarction, coronary re-
vascularization procedure, ischemic stroke, peripheral
arterial disease and CVD death was used as the primary
outcome. CVD death and all-cause mortality were sec-
ondary outcomes. Adjusted Cox proportional hazards
models were used to compare primary and secondary
outcomes across diet drink intake strata.
KEY RESULTS: In all, 59,614 women, mean age
62.8 years, were included for analysis. In unadjusted
analysis over a follow-up of 8.7±2.7 years, the primary
outcome occurred in 8.5 % of the women consuming ≥2
diet drinks/day, compared to 6.9 %, 6.8 % and 7.2 % in
the 5–7/week, 1–4/week and 0–3/month groups, respec-
tively. After controlling for other CVD risk factors, women
who consumed ≥2 drinks/day had a higher adjusted risk
of CVD events (HR 1.3, 95 % CI 1.1–1.5), CVD mortality
(HR 1.5, 95 % CI 1.03–2.3) and overall mortality (HR 1.3,
95 % CI 1.04–1.5) compared to the reference group (0–3
CONCLUSIONS: This analysis demonstrates an associa-
tion between high diet drink intake and CVD outcomes
and mortality in post-menopausal women in the WHI OS.
KEY WORDS: diet; cardiovascular diseases; lifestyle; diet drinks; artificial
J Gen Intern Med
© Society of General Internal Medicine 2014
Observational studies have demonstrated an association be-
tween sugar-sweetened beverages and obesity, metabolic syn-
drome, and cardiovascular disease (CVD) outcomes; this as-
sociation is attributed to increased energy and sugar load.
Diet drinks are often consumed as a low-calorie alternative to
sugar-sweetened drinks, and one out of every five Americans
drinks diet soda every day.
Several population-based studies
have demonstrated a positive association between diet drinks
and the metabolic syndrome, which in turn is associated with
increased risk for CVD.
However, data are limited as to
whether there is an increased CVD risk associated with diet
drink consumption. Given the large population directly affect-
ed by such an association, this study was performed to eval-
uate the effect of diet drink consumption on CVD outcomes in
the Women’s Health Initiative Observational Study (WHI OS)
The WHI is a multicenter national study that involved 40
centers across 24 states and the District of Columbia and
included 161,808 postmenopausal women.
of three overlapping Clinical Trials (CT) including 68,133
women, as well as an Observational Study (OS) enrolling
93,676 subjects. Recruitment to the WHI took place from
1993 to 1998, and enrolled postmenopausal women between
50 and 79 years of age.
The OS cohort was derived from the
women who were screened for the clinical trials but were
ineligible or unwilling to be randomized, as well as from those
Electronic supplementary material The online version of this article
(doi:10.1007/s11606-014-3098-0) contains supplementary material,
which is available to authorized users.
Received April 21, 2014
Revised August 11, 2014
Accepted October 27, 2014
who responded to a direct invitation to be screened for the
Exclusion criteria (for both the CT and the OS) were the
presence of any medical condition with a predicted survival of
< 3 years, concerns about adherence, and active participation
in other randomized trials. An effort was made to enroll
women of racial and ethnic minority groups, with a target of
20 % of overall enrollment.
No significant sampling issues
were noted with regards to participant enrollment. All partic-
ipants provided written informed consent, and the study pro-
tocol was approved by institutional review boards of all par-
The OS cohort was used for analysis for this study. After
baseline evaluation at a clinic visit, updates regarding medical
histories and selected exposure data were obtained annually by
mailed questionnaire. All participants were also invited to a
clinic follow-up visit at 3 years after enrollment.
information was collected using food frequency question-
naires obtained at baseline and follow-up clinic visits.
questionnaire at follow-up year 3 also included information
regarding diet drink consumption over the previous 3 months.
Diet drink intake was categorized into nine groups on the
questionnaire, ranging from none or less than one per month
to more than six drinks per day, with each drink defined as the
equivalent of a 12-oz can of beverage. Follow-up ended in
2005, and ranged from six to ten years, depending on year of
Inclusion and Exclusion Criteria
All women in the OS for whom diet drink intake data was
available were included for analysis in this study. Exclusion
criteria were (1) presence of any of the following pre-existing
diagnoses: coronary heart disease (CHD), heart failure (HF),
myocardial infarction (MI), coronary revascularization proce-
dure, ischemic stroke, peripheral arterial disease (PAD) as well
as other related cardiovascular and thromboembolic events
(angina, carotid artery disease, hemorrhagic stroke, transient
ischemic attack, pulmonary embolism and deep vein throm-
bosis); or (2) a survival of less than 60 days after collection of
diet drink consumption data.
A composite of incident CHD, HF, MI, coronary revascular-
ization procedure, ischemic stroke, PAD and CVD death was
used as the primary outcome, with time to first occurrence of
any of these events being the relevant endpoint. Secondary
outcomes were rates of CVD death, as well as death from any
cause. Outcomes in the OS were identified through self-report
at annual contacts, and specific details were obtained as need-
ed using standardized questionnaires and request of medical
Data linkage with the National Death Index was
performed to assure completeness of survival data, and causes
of death were actively investigated. Adjudication of outcomes
was performed by physicians at the Clinical Centers, the
Clinical Coordinating Center, and the National Institutes of
Health, using a staged approach. All primary and safety end-
points, as well as a sample of locally adjudicated secondary
endpoints, were reviewed centrally. The adjudicating physi-
cians were blinded to any participant information that could
potentially result in bias.
Risk factors and demographic characteristics included base-
line self-reported age, race, education, income, body mass
index (BMI), smoking, alcohol intake, hormone therapy
(HT) use, physical activity, energy intake, salt intake, history
hypertension, or high cholesterol, as well as
sugar-sweetened beverage intake. Age was assessed in years
as both continuous and categorical (49–59, 60–64, 65–69, ≥
70), physical activity was measured as total energy expended,
and energy intake was measured as calibrated total calories.
Unadjusted relationships between continuous baseline var-
iables were assessed using t-tests comparing least square
means from general linear models, and between categorical
variables using the Pearson chi-square test. Occurrence of a
CVD event, CVD death and all-cause mortality were mea-
sured in years from the date of Visit 3 until the first CVD
event, death or until the participant was last known to be alive.
Unadjusted differences in incidence of primary and secondary
outcomes across diet drink consumption strata were assessed
using the Pearson chi-square test. Cox proportional hazards
models were used to assess the relationship of diet soda intake
with occurrence of the first CVD event or time to death,
adjusting for CVD risk factors.
Hazard ratios (HR) and
95 % confidence intervals (CI) were calculated to measure
the magnitude of the associations. Four different models were
constructed. Model I was unadjusted; Model II was adjusted
for age, race, education and income; Model III was adjusted
for the variables in Model II plus smoking status, BMI, and a
history of diabetes, hypertension and hyperlipidemia; and
Model IV was adjusted for the variables in Model III plus
alcohol intake, log calibrated energy intake, physical activity,
sugar-sweetened beverage intake, salt intake, and HT. Subjects
with missing variables were excluded from analysis. All var-
iables were assessed for the proportional hazards assumption
before inclusion in the model. Adjusted cardiac event models
were stratified on hypertension and high cholesterol, and
adjusted mortality models were stratified on hypertension
and the history of diabetes, because these variables did not
meet the proportional hazards assumption.
Additional sensitivity and exploratory analyses were also
performed. In order to decrease the risk of reverse causality,
women with a history of diabetes, hypertension, and high
cholesterol were excluded and the adjusted risk (Model IV
above) for the primary outcome was calculated in the residual
healthier population. The impact of missing salt intake and
calibrated energy intake data was evaluated by constructing
fully adjusted models for the primary outcome after excluding
these variables. The effect of BMI on the relationship between
relationship of diet drink intake and outcomes was further
investigated by using interaction terms for BMI and diet drink
Vyas et al.: Diet Drinks and Cardiovascular Events JGIM
categories in the fully adjusted models. The role of diet quality
was assessed by including Healthy Eating Index-2005 (HEI-
2005) scores in the model (with salt and energy intake exclud-
ed to prevent double counting). HEI-2005 has previously been
used in the WHI and other population based cohorts, and has
been demonstrated to be associated with both cardiometabolic
risk factors and cardiovascular outcomes and mortality.
Finally, as the definition of HF in the WHI was sub-optimal, a
revised primary outcome that excluded HF was constructed,
and adjusted hazard ratios were calculated for the revised
primary outcome as well as HF alone.
This manuscript was prepared in accordance with the
Strengthening the Reporting of Observational Studies in Epi-
demiology (STROBE) Statement.
All statistical significance
was based on two-tailed tests and p values ≤0.050. Statistical
analyses were performed using SAS 9.3 [SAS System for
Windows, version 9.3. Cary, NC: SAS Institute. 2002–2010].
Of 93,676 women who were part of the WHI OS, 59,614 met
criteria for this study and were included for analysis (Fig. 1).
Among non-deceased women who met other inclusion
criteria, 9,946 were excluded due to lack of diet drink data
secondary to either absent or partial year 3 follow-up. Those
who were excluded had a higher proportion of women older
than 70 years of age as well as women of African American or
Hispanic race, and had lower income and educational status,
compared to women who were included in the analysis.
Almost two-thirds of those included (38,337 women,
64.3 %) fell into the lowest consumption category (0–3 drinks
per month), and there were 11,590 (19.4 %), 6,702 (11.2 %)
and 2,985 (5.0 %) women who consumed 1–4 drinks per
week, 5–7 drinks per week, and two or more drinks per day,
respectively (Table 1). Categorization of diet drink intake was
constructed in accordance with previous studies and in order to
satisfy proportional hazard assumptions.
The mean fol-
low up was 8.7 (SD±2.7) years. The average age of the
included cohort was 62.8 (SD±7.2) years, and a majority of
women (85.7 %) were white.
Baseline variables differed significantly between the four
groups (Table 1). Women who consumed ≥2 diet drinks/day
were younger than the other groups, with a mean age of
59.5 years, compared to 63.5 years in the 0–3 drinks per month
group. There was increased prevalence of diabetes and hyper-
tension in the women who consumed the most diet drinks, and
they had a higher BMI, a greater proportion of smokers and
higher calibrated energy intake. When non-diet drink con-
sumption was assessed across the different diet drink intake
strata, no significant difference was found.
The composite primary outcome occurred in 8.5 % of the
women who consumed ≥2 diet drinks/day, compared to 6.9,
6.8 and 7.2 % of the women in the 5–7 drinks/week, 1–4
drinks/week and 0–3 drinks/month groups, respectively
(Table 2). Unadjusted incidence of the secondary out-
comes did not differ markedly between the ≥2diet
drinks/day and 0–3 diet drinks/month groups, with inci-
dence of CVD death and overall mortality being 1.6 and
7.8 %, respectively, in the highest consumption group, and
1.7 and 7.7 % in the lowest (Table 2).
Cox proportional hazard models demonstrated that women
who consumed ≥2 drinks/day had a higher unadjusted risk of
experiencing CVD events (in the form of the combined end-
point) as compared to those who consumed 0–3 drinks/month
(the reference group), with the difference being statistically
significant (Model I; N: 59,614; HR 1.2, 95 % CI 1.04–1.3)
(Table 3). This increased risk among the highest consumption
category persisted when the models were adjusted progres-
sively for baseline demographic variables (Model II; N:
55,073; HR 1.6, 95 % CI 1.4–1.9), common CVD risk factors
(Model III; N: 53,037; HR 1.3. 95 % CI 1.2–1.5), and other
possible confounders (Model IV; N: 33,619; HR 1.3, 95 % CI
1.1–1.5). No significant difference in risk was found between
the other consumption groups and the reference group, with
regards to the primary endpoint. There was progressive de-
crease in sample size with the Cox proportional hazard models
secondary to missing data, with Model IV especially limited
by missing values for energy and salt intake.
Unadjusted analysis (Model I) did not demonstrate a differ-
ence between the ≥2 drinks/day group and the 0–3 drinks/
month group with regards to either CVD death (N: 59,447; HR
0.9, 95 % CI 0.7–1.2) or overall mortality (N: 59,447; HR 1.0,
95 % CI 0.9–1.1) (Table 3). However, with adjustment for
baseline demographic characteristics, CVD risk factors and
other confounders, there emerged a significant risk of both
CVD death (Model II: N 55,073, HR 1.8, 95 % CI 1.3–2.4;
Model III: N 53,037, HR 1.4, 95 % CI 1.04–1.9; Model IV: N
33,619, HR 1.5, 95 % CI 1.03–2.3) and overall mortality
(Model II: N 55,073, HR 1.5, 95 % CI 1.3–1.7; Model III: N
53,037, HR 1.4, 95 % CI 1.2–1.6; Model IV: N 33,619, HR
1.3, 95 % CI 1.04–1.5) with consumption of ≥2 diet drinks/
day. Hazard ratios for adjustment covariates for both primary
and secondary outcomes are listed in the online appendix.
Figure 1 Selection of study cohort.
Vyas et al.: Diet Drinks and Cardiovascular EventsJGIM
Sensitivity and Exploratory Analyses
A history of diabetes, hypertension and high cholesterol was
absent in 38,658 women, and these were classified as being
healthier at baseline. The hazard ratios for the primary outcome
for these were similar to those for the primary analysis;
however, the 95 % confidence intervals were wider and did
not reach statistical significance (Table 4). Salt and calibrated
energy intake data were missing in 18,097 (30.4 %) women,
and this was uniformly distributed across the diet drink con-
sumption groups. Excluding salt and calibrated energy intake
from the model did not result in a significant change in the
Table 1. Baseline Characteristics of Women Included in Analysis
Characteristics Diet drink consumption
0–3/month 1–4/week 5–7/week ≥2/day p value*
N (%) 38,337 (64.3) 11,590 (19.4) 6,702 (11.2) 2,985 (5.0)
Age (Mean ± SD) 63.5±7.2 62.3±6.9 61.2±6.9 59.5±6.5 < 0.0001
Race (%) < 0.0001
American Indian 128 (0.3) 28 (0.2) 30 (0.4) 13 (0.4)
Asian/Pacific Islander 1,401 (3.7) 270 (2.3) 130 (1.9) 36 (1.2)
African American 2,559 (6.7) 720 (6.2) 387 (5.8) 207 (7.0)
Hispanic 1,282 (3.4) 363 (3.1) 207 (3.1) 98 (3.3)
White 32,418 (84.8) 10,068 (87.1) 5880 (88.0) 2601 (87.4)
Other 441 (1.2) 109 (0.9) 49 (0.7) 22 (0.7)
Education (%) < 0.0001
< High school, High school 10,498 (27.6) 3,426 (29.8) 1,936 (29.1) 893 (30.2)
Some college 9,905 (26.0) 3,009 (26.2) 1,808 (27.2) 838 (28.3)
College graduate, > College 17,656 (46.4) 5,069 (44.1) 2,899 (43.6) 1,228 (41.5)
Income in dollars (%) < 0.0001
< 35,000 12,759 (35.8) 3,631 (33.5) 1,902 (30.2) 925 (32.9)
35,000–< 75,000 14,852 (41.7) 4,643 (42.9) 2,772 (44.0) 1,193 (42.4)
≥75,000 8,009 (22.5) 2,551 (23.6) 1,628 (25.8) 693 (24.7)
BMI (%) < 0.0001
17,692 (46.5) 3,882 (33.7) 1,986 (29.9) 686 (23.2)
12,765 (33.6) 4,327 (37.6) 2,424 (36.5) 1,028 (34.7)
7,572 (19.9) 3,301 (28.7) 2,234 (33.6) 1,245 (42.1)
Smokers (%) 1716 (4.5) 394 (3.4) 303 (4.5) 227 (7.7) < 0.0001
Alcohol intake (%) 0.086
< 0.5 drinks/week 22,086 (57.7) 6,812 (58.9) 3,875 (57.9) 1,769 (59.3)
0.5–2 drinks/week 5,821 (15.2) 1,684 (14.6) 1,001 (15.0) 467 (15.7)
> 2 drinks/week 10,352 (27.1) 3,077 (26.6) 1,812 (27.1) 745 (25.0)
HT users (%) < 0.0001
Never 12,548 (32.8) 3369 (29.1) 1896 (28.3) 878 (29.6)
Estrogen only 11,014 (28.8) 3623 (31.3) 2072 (31.0) 943 (31.7)
Either one or both 14,565 (38.1) 4541 (39.3) 2697 (40.3) 1136 (38.2)
Unknown 129 (0.3) 33 (0.3) 29 (0.4) 14 (0.5)
Physical activity (total energy expended, MET-hours/week, Mean ± SD) 12.0±13.4 12.1±13.4 11.6±13.0 11.9±13.2 0.38
Calibrated energy intake (kcal, Mean ± SD) 2033±187 2092±196 2130±209 2190±232 < 0.0001
Salt intake (mg, Mean ± SD) 2393±1077 2519±1098 2601±1193 2732±1388 < 0.0001
History of diabetes (%) 2636 (6.9) 1178 (10.2) 810 (12.1) 502 (16.8) < 0.0001
History of hypertension (%) 9962 (26.2) 3281 (28.5) 1956 (29.4) 927 (31.2) < 0.0001
History of high cholesterol (%) 4264 (11.4) 1512 (13.3) 844 (12.8) 384 (13.1) < 0.0001
Sugar-sweetened beverage intake 0.44
None or < 1/month 23,794 (62.1) 7,111 (61.4) 4,133 (61.7) 1,831 (61.3)
< Once a day 12,798 (33.4) 3,944 (34.0) 2,253 (33.6) 1,012 (33.9)
Daily 1,745 (4.6) 535 (4.6) 316 (4.7) 142 (4.8)
SD standard deviation, BMI body mass index, HT hormone therapy
*Unadjusted p value
Table 2. Incidence of Primary and Secondary Outcomes Across Diet Drink Consumption Strata
Outcomes Diet drink consumption
0–3/month 1–4/week 5–7/week ≥2/day p value*
Total subjects 38,337 11,590 6,702 2,985
Cardiovascular events (%)
Combined 2,745 (7.2) 785 (6.8) 462 (6.9) 254 (8.5) 0.010
CHD 1,030 (2.7) 279 (2.4) 159 (2.4) 94 (3.1) 0.055
HF 349 (0.9) 95 (0.8) 53 (0.8) 37 (1.2) 0.13
MI 819 (2.1) 219 (1.9) 122 (1.8) 74 (2.5) 0.065
Coronary revascularization 1,039 (2.7) 325 (2.8) 204 (3.0) 111 (3.7) 0.008
Stroke 652 (1.7) 192 (1.7) 94 (1.4) 58 (1.9) 0.212
PAD 160 (0.4) 49 (0.4) 27 (0.4) 22 (0.7) 0.079
Cardiac death 642 (1.7) 163 (1.4) 90 (1.3) 47 (1.6) 0.076
All cause death 2,970 (7.7) 789 (6.8) 445 (6.6) 233 (7.8) 0.003
CHD coronary heart disease, HF heart failure, MI myocardial infarction, PAD peripheral arterial disease
*Unadjusted p values (χ
Vyas et al.: Diet Drinks and Cardiovascular Events JGIM
hazard ratios for the primary outcome (HR for ≥2 diet drink/
daygroup:1.3;95%CI1.1–1.5). Addition of the HEI-2005
score in the model also did not lead to any notable change in the
association between diet drink consumption and CVD events
(HR for ≥2 diet drinks/day group: 1.3; 95 % CI 1.1–1.5).
A significant unadjusted and adjusted interaction was ob-
served between baseline BMI and diet drink consumption for
the primary outcome (p=0.003 for both) (Online appendix).
On adjusted analysis, consumption of ≥2 diet drinks/day
continued to be significantly associated with occurrence of
CVD events in women with BMI ≥30 (HR 1.3, 95 % CI 1.1–
1.6) and BMI < 25 (HR 1.7, 95 % CI 1.3–2.4), while the
association was not statistically significant in women with
BMI 25–30 (HR 1.2, 95 % CI 0.9–1.5).
Finally, exclusion of HF from the primary outcome did not
change the hazard ratios significantly compared to the primary
analysis, and these continued to be statistically significant for
the ≥2 diet drinks/day group (Online appendix).
This study demonstrates an association between regular daily
intake of two or more diet drinks and CVD outcomes and
mortality in post-menopausal women. In our analysis, women
who consumed ≥2 diet drinks/day had a 30 % higher adjusted
risk of CVD events as well as overall mortality compared to
those with an intake of 0–3 diet drinks/month.
Our study, with nearly 60,000 subjects, is one of the largest
studies evaluating diet drink consumption and outcomes. Our
findings are concordant with the results of the only previous
report that expressly evaluated the risk of CVD events with diet
Gardener et al. used the Northern Manhattan
Study (NOMAS) cohort, and included 2,564 subjects (64 %
female). They found daily diet drink consumption to be asso-
ciated with an increased adjusted risk of CVD events (incident
stroke, MI or vascular death). The overall event rate was about
23 % over a mean follow-up of 9.8 years, which is significantly
higher than the total event rate of 7.1 % over 8.7 years in our
study; some of this may be explained by the presence of both
men and women in the NOMAS subject population, as well as
more ethnic diversity compared to the WHI cohort.
The results are also supported by other observational data
that have shown a link between diet drink consumption and
metabolic syndrome. Three separate reports, involving pa-
tients from the Framingham Heart Study, the Atherosclero-
sis Risk in Communities study and the Multi-Ethnic Study
of Atherosclerosis demonstrated significantly increased
rates of incident metabolic syndrome among subjects who
were frequent consumers of diet drinks.
that analyzed data from the San Antonio Heart Study noted a
greater incidence of obesity and a significantly higher
Table 3. Hazard Ratios for Primary and Secondary Outcomes
Across Diet Drink Consumption Strata
Cardiovascular events Cardiac death Overall death
HR (95 % CI) HR (95 % CI) HR (95 % CI)
Model I* N: 59,614 N: 59,447 N: 59,447
0–3/month 1.0 1.0 1.0
1–4/week 0.9 (0.9–1.02) 0.8 (0.7–0.99) 0.9 (0.8–0.9)
5–7/week 1.0 (0.9–1.1) 0.8 (0.6–0.99) 0.8 (0.8–0.9)
≥2/day 1.2 (1.04–1.3) 0.9 (0.7–1.2) 1.0 (0.9–1.1)
N: 55,073 N: 55,073 N: 55,073
0–3/month 1.0 1.0 1.0
1–4/week 1.0 (0.96–1.1) 1.0 (0.8–1.2) 1.0 (0.9–1.1)
5–7/week 1.1 (1.0–1.3) 1.1 (0.9–1.4) 1.1 (1.0–1.2)
≥2/day 1.6 (1.4–1.9) 1.8 (1.3–2.4) 1.5 (1.3–1.7)
N: 53,037 N: 53,037 N: 53,037
0–3/month 1.0 1.0 1.0
1–4/week 1.0 (0.9–1.1) 0.9 (0.8–1.1) 1.0 (0.9–1.1)
5–7/week 1.0 (0.9–1.1) 1.0 (0.8–1.3) 1.1 (1.0–1.2)
≥2/day 1.3 (1.2–1.5) 1.4 (1.04–1.9) 1.4 (1.2–1.6)
0–3/month 1.0 1.0 1.0
1–4/week 1.0 (0.9–1.1) 0.9 (0.7–1.2) 1.0 (0.9–1.1)
5–7/week 1.1 (0.9–1.2) 0.9 (0.7–1.3) 1.1 (0.9–1.2)
≥2/day 1.3 (1.1–1.5) 1.5 (1.03–2.3) 1.3 (1.04–1.5)
HR hazard ratios, CI confidence intervals
*Model I was unadjusted
Model II was adjusted for age, race, education and income
Model III was adjusted for the variables in Model II plus smoking
status, BMI, and a history of diabetes, hypertension and hyperlipidemia
Model IV was adjusted for the variables in Model III plus alcohol
intake, log calibrated energy intake, physical activity, sugar-sweetened
beverage intake, salt intake, and hormone therapy. Adjusted cardiac
event models were stratified on hypertension, and high cholesterol and
adjusted mortality models were stratified on hypertension and the
history of diabetes because these variables did not meet the proportional
The majority of loss of patients in model 4 was secondary to missing
data for salt intake and calibrated energy
Table 4. Adjusted Hazard Ratios for Primary Outcome Across Diet
Drink Consumption Strata for Primary Analysis, for Baseline
Healthy Women, with Salt Intake and Calibrated Energy Intake
Excluded, and with Healthy Eating Index-2005 Included
Diet drink consumption
(95 % CI)
(95 % CI)
(95 % CI)
(95 % CI)
N: 33,619 1.0
1.0 (0.9–1.1)1.1 (0.9–1.2)1.3 (1.1–1.5)
1.0 (0.9–1.2)1.0 (0.8–1.2)1.2 (0.9–1.7)
Salt and calibrated
1.0 (0.9–1.1)1.0 (0.9–1.1)1.3 (1.1–1.5)
1.0 (0.9–1.1)1.0 (0.9–1.2)1.3 (1.1–1.5)
HR hazard ratio, CI confidence interval
*Model IV: Adjusted for age, race, education, income, smoking status,
BMI, diabetes, hypertension, hyperlipidemia, alcohol intake, log
calibrated energy intake, physical activity, sugar-sweetened beverage
intake, salt intake, and hormone therapy
Model IV; Included women without a history of diabetes, hypertension,
and high cholesterol
Adjusted for Model IV excluding salt intake and calibrated energy
Adjusted for Model IV including Healthy Eating Index-2005 scores and
excluding salt intake and calibrated energy intake
Vyas et al.: Diet Drinks and Cardiovascular EventsJGIM
increase in BMI with increasing intake of artificially sweet-
ened beverages, with an apparent dose response relationship
between the amount of artificially sweetened beverages
consumed and weight gain.
As both metabolic syndrome
and obesity are important risk factors for CVD, this may
contribute in part to the higher incidence of CVD events in
Exact pathophysiologic mechanisms that would explain the
association of weight gain, development of metabolic syn-
drome, and increased CVD events with diet soda consumption
are still unclear. One hypothesis is that artificial sweeteners
may increase the desire for sugar-sweetened, energy-dense
beverages/foods. Experimental data from animal (rat) models
suggests that consumption of products containing artificial
sweeteners may disrupt the correlation between sweet taste
and the energy content of foods (thus interfering with funda-
mental homeostatic and physiological processes).
study investigating the functional magnetic resonance imaging
response to sucrose (a nutritive sweetener) and saccharin (a
nonnutritive sweetener) in diet soda drinkers versus non-diet
soda drinkers found alterations in the reward processing of
sweet taste in individuals who regularly consume diet soda.
An alternative explanation for this association could be
confounding by dietary patterns or incomplete adjustment
for confounders. In a recent study, diet beverage consumers
defined as having a healthy diet had a lower risk of metabolic
Analysis from the Health Professionals Follow-
Up Study presented evidence of an association between con-
sumption of sugar-sweetened beverages and increased CHD
risk and intermediate biomarkers, but no associations were
found for artificially sweetened beverage intake.
the study noted that artificially sweetened beverage consump-
tion was associated with healthy lifestyle traits and higher
overall diet quality, which may suggest a role of dietary pattern
in the determination of outcomes. In addition, individuals
attempting to restrict energy intake and control weight may
be more likely to consider artificially sweetened beverages, a
factor that may influence CVD risk factor associations.
sensitivity analysis evaluating women without diabetes, hy-
pertension and high cholesterol at baseline was done to ad-
dress some of these questions. While the confidence intervals
widened (with loss of statistical significance), the hazard ratios
did not change significantly. Additionally, inclusion of the
HEI-2005 score, which is a validated tool for quantifying
dietary quality, also did not result in a significant change in
the association between diet drink consumption and CVD
events. Finally, the interaction between BMI and diet drink
intake noted in our analysis is an interesting finding that bears
further investigation, especially given other recent data that
highlight the role of body weight in this setting.
Limitations of this study include its observational nature,
and the fact that it involved retrospective analysis of data not
collected expressly for the purpose of this paper. It also in-
volves a specific population, that of post-menopausal women,
and thus may not be generalizable to other populations. As the
differences in demographic characteristics between women
excluded for missing diet drink data and the included women
demonstrate, presence of selection bias cannot be ruled out.
Due to limited data collection regarding diet drink intake after
the baseline evaluation, the analysis does not take into account
changes in consumption pattern over the course of follow-up.
Finally, the association seen in this study does not translate
into causality, and may be due to confounding variables that
were not examined in multivariate analysis, or may not even
be clearly defined as being confounders in this relationship as
In conclusion, this study suggests an association between
consumption of two or more diet drinks per day and adverse
CVD events, as well as increased mortality. However, further
evaluation with other clinical studies, animal models and even
molecular and pharmacologic analyses is needed to confirm or
disprove this link, and to assess a possible causal relationship
between diet drink intake and increased CVD risk.
Funding Sources: The Division of Cardiovascular Medicine at the
University of Iowa Hospitals & Clinics and the Women’s Health Initia-
tive provided partial funding for this manuscript.
Conflict of Interest: The authors declare that they do not have a
conflict of interest.
Relationship with Industry: None.
Corresponding Author: Ankur Vyas, MD; Division of Cardiovascular
MedicineUniversity of Iowa Hospitals Clinics, 200 Hawkins Dr., Int. Med.
E316-1 GH, Iowa City, IA 52242, USA (e-mail: firstname.lastname@example.org).
1. Fung TT, Malik V, Rexrode KM, Manson JE, Willett WC, Hu FB. Sweetened
beverage consumption and risk of coronary heart disease in women. Am J
Clin Nutr. 2009;89(4):1037–1042. doi:10.3945/a jcn.2008. 27140.
2. Dhingra R, Sullivan L, Jacques PF, et al. Soft drink consumption and
risk of developing cardiometabolic risk factors and the metabolic syndrome
in middle-aged adults in the community. Circulation. 2007;116(5):480–
3. Malik VS, Popkin BM, Bray GA, Després J-P, Hu FB. Sugar-sweetened
beverages, obesity, type 2 diabetes mellitus, and cardiovascular disease
risk. Circulation. 2010;121(11):1356–1364. doi:10.1161/
4. Qi Q, Chu AY, Kang JH, et al. Sugar-sweetened beverages and genetic risk
of obesity. N Engl J Med. 2012;367(15):1387–1396. doi:10.1056/
5. Schulze MB, Manson JE, Ludwig DS, et al. Sugar-sweetened beverages,
weight gain, and incidence of type 2 diabetes in young and middle-aged
women. JAMA. 2004;292(8):927–934. doi:10.1001/jama.292.8.927.
6. Fakhouri THI, Kit BK, Ogden CL. Consumption of diet drinks in the
United States, 2009–2010. NCHS Data Brief. 2012;109:1–8.
7. Fowler SP, Williams K, Resendez RG, Hunt KJ, Hazuda HP, Stern MP. Fueling
the obesity epidemic? Artificially sweetened beverage use and long-term
weight gain. Obesity. 2008;16(8):1894–1900. doi:10.1038/oby.2008.284.
8. Nettleton JA, Lutsey PL, Wang Y, Lima JA, Michos ED, Jacobs DR. Diet
soda intake and risk of incident metabolic syndrome and type 2 diabetes in
the Multi-E thnic Study of Atherosclerosis (MESA). Diabetes Care.
9. Lutsey PL, Steffen LM, Stevens J. Dietary intake and the development of the
metabolic syndrome: the atherosclerosis risk in communities study. Circulation.
Vyas et al.: Diet Drinks and Cardiovascular Events JGIM
10. Rossouw JE, Hurd S. The Women’s Health Initiative: recruitment
complete–looking back and looking forward. J Womens Health.
11. Hays J, Hunt JR, Hubbell FA, et al. The Women’s Health Initiative
recruitment methods and results. Ann Epidemiol. 2003;13(9 Suppl):S18–
12. Langer RD, White E, Lewis CE, Kotchen JM, Hendrix SL, Trevisan M.
The Women’s Health Initiative Observational Study: baseline charac-
teristics of participants and reliability of baseline measures. Ann
Epidemiol. 2003;13(9 Suppl):S107–21. doi:10.1016/S1047-
13. The Women’s Health Initiative Study Group. Design of the Women’s Health
Initiative clinical trial and observational study. Control Clin Trials.
14. Patterson RE, Kristal AR, Tinker LF, Carter RA, Bolton MP, Agurs-
Collins T. Measurement characteristics of the Women’s Health Initiative
food frequency questionnaire. Ann Epidemiol. 1999;9(3):178–187.
15. Curb JD, Mctiernan A, Heckbert SR, et al. Outcomes ascertainment and
adjudication methods in the Women’s Health Initiative. Ann Epidemiol.
2003;13(9 Suppl):S122–8. doi:10.1016/S1047-2797(03)00048-6.
16. Margolis KL,Lihong Q, Brzyski R, etal. Validity of diabetes self-reports in
the Women’s Health Initiative: comparison with medication inventories and
fasting glucose measurements. Clin Trials. 2008;5(3):240–247.
17. Neuhouser ML,Tinker L, Shaw PA, et al. Use of recovery biomarkers t o
calibrate nutrient consumption self-reports in the Women’sHealth
Initiative. Am J Epidemiol. 2008;167(10):1247–1259. doi:10.1093/
18. Cox DR. Regression models and life-tables. J R S tat Soc Ser B.
19. Nicklas TA, O’Neil CE, Fulgoni VL. Diet quality is inversely related to
cardiovascular risk factors in adults. J Nutr. 2012;142(12):2112–2118.
20. Chiuve SE, Fung TT, Rimm EB, et al. Alternative dietary indices both
strongly predict risk of chronic disease. J Nutr. 2012;142(6):1009–1018.
21. George SM, Ballard-Barbash R, Shikany JM, et al. Better postdiagnosis
diet quality is associated with reduced risk of death among postmeno-
pausal women with invasive breast cancer in the Women’s Health Initiative.
Cancer Epidemiol Biomark Prev. 2014;23(4):575–583. doi:10.1158/1055-
22. Guenther PM, Reedy J, Krebs-Smith SM. Development of the healthy
eating index-2005. J Am Diet Assoc. 2008;108(11):1896–1901.
23. Guenther PM, Reedy J, Krebs-Smith SM, Reeve BB. Evaluation of the
healthy eating index-2005. J Am Diet A ssoc. 2008;108(11):1854–1864.
24. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC,
Van d en br o uck e J P. The Strengthening the Reporting of Observational
Studies in Epidemiology (STROBE) statement: guidelines for reporting
observational studies. Ann Intern Med. 2007;147(8):573–577.
25. Gardener H, Rundek T, Markert M, Wright CB, Elkind MSV, Sacco RL.
Diet soft drink consumption is associated with an increased risk of
vascular events in the Northern Manhattan Study. J Gen Intern Med.
2012;27(9):1120–1126. d oi: 10.1007/s11606-011-1968-2.
26. Swithers SE, Davidson TL. A role for sweet taste: calorie predictive
relations in energy regulation by rats. Behav Neurosci. 2008;122(1):161–
27. Green E, Murphy C. Altered processing of sweet taste in the brain of diet
soda drinkers. Physiol Behav. 2012;107(4):560–567. doi:10.1016/
28. Duffey KJ, Steffen LM, Van Horn L, Jacobs DR, Popkin BM. Dietary
patterns matter: diet beverages and cardiometabolic risks in the
longitudinal Coronary Artery Risk Development in Young Adults
(CARDIA) Study. Am J Clin Nutr. 2012;95(4):909–915. doi:10.3945/
29. de Koning L, Malik VS, Kellogg MD, Rimm EB, Willett WC, Hu FB.
Sweetened beverage consumption, incident coronary heart disease, and
biomarkers of risk in men. Circula tion. 2012;125(14):17 35–41-S1.
30. Elfhag K, Tynelius P, Rasmussen F. Sugar-sweetened and artificially
sweetened soft drinks in association to restrained, external and emotional
eating. Physiol Behav. 2007;91(2–3):191–195. doi:10.1016/
31. Bleich SN, Wolfson JA, Vine S, Wang YC. Diet-beverage consumption and
caloric intake among US adults, overall and by body weight. Am J Public
Health. 2014:e1–e7. doi:10.2105/AJPH.2013.301556.
Vyas et al.: Diet Drinks and Cardiovascular EventsJGIM