1390 Diabetes Care Volume 42, August 2019

F. Javier Basterra-Gortari,1,2
Effects of a Mediterranean Eating Miguel Ruiz-Canela,1,3
CLIN CARE/EDUCATION/NUTRITION/PSYCHOSOCIAL

Miguel A. Martı́nez-González,1,3,4
Plan on the Need for Glucose- Nancy Babio,3,5 José V. Sorlı́,3,6
Montserrat Fito,3,7 Emilio Ros,3,8
Lowering Medications in Enrique Gómez-Gracia,3,9 Miquel Fiol,3,10
José Lapetra,3,11 Ramón Estruch,3,12
Participants With Type 2 Diabetes: Luis Serra-Majem,3,13 Xavier Pinto,3,14
José I. González,3,6 Mónica Bulló,3,5
A Subgroup Analysis of the Olga Casta~ ner,3,7 Ángel Alonso-Gómez,3,15
Luis Forga, and Fernando Arós,3,15,17
16

PREDIMED Trial for the PREDIMED Study Investigators

Diabetes Care 2019;42:1390–1397 | https://doi.org/10.2337/dc18-2475

1
Department of Preventive Medicine and Public
Health, University of Navarra, Pamplona, Spain
2
Department of Internal Medicine (Endocrinology),
Hospital Reina Sofia, Tudela, Spain
3
Consorcio CIBER, M.P. Fisiopatologı́a de la Obe-
sidad y Nutrición (CIBERObn), Instituto de Salud
Carlos III (ISCIII), Madrid, Spain
4
Department of Nutrition, Harvard T.H. Chan
School of Public Health, Boston, MA
5
OBJECTIVE Universitat Rovira i Virgili, Department Bio-
chemistry and Biotechnology, Human Nutrition
To examine the effects of two Mediterranean eating plans (Med-EatPlans) versus a Unit, IISPV, Reus, Spain
6
low-fat eating plan on the need for glucose-lowering medications. Department of Preventive Medicine, University
of Valencia, Valencia, Spain
7
RESEARCH DESIGN AND METHODS Cardiovascular Risk and Nutrition (Regicor Study
Group), Hospital del Mar Medical Research In-
From the Prevención con Dieta Mediterránea (PREDIMED) trial, we selected 3,230 stitute, Barcelona, Spain
participants with type 2 diabetes at baseline. These participants were randomly 8
Lipid Clinic, Endocrinology and Nutrition Service,
assigned to one of three eating plans: Med-EatPlan supplemented with extra- Institut d’investigacions Biomèdiques August Pi i
virgin olive oil (EVOO), Med-EatPlan supplemented with mixed nuts, or a low-fat Sunyer (IDIBAPS), Hospital Clı́nic, University of
Barcelona, Barcelona, Spain
eating plan (control). In a subgroup (15%), the allocation was done in small 9
Department of Preventive Medicine, University
clusters instead of using individual randomization, and the clustering effect was of Malaga, Malaga, Spain
10
taken into account in the statistical analysis. In multivariable time-to-event survival Institute of Health Sciences, University of Ba-
models, we assessed two outcomes: 1) introduction of the first glucose-lowering learic Islands and Son Espases Hospital, Palma
de Mallorca, Spain
medication (oral or injectable) among participants on lifestyle management at 11
Department of Family Medicine, Research
enrollment and 2) insulin initiation. Unit, Distrito Sanitario Atención Primaria Sevilla,
Sevilla, Spain
12
RESULTS Department of Internal Medicine, Institut
After a median follow-up of 3.2 years, in multivariable analyses adjusting for d’investigacions Biomèdiques August Pi i Sunyer
(IDIBAPS), Hospital Clı́nic, University of Barce-
baseline characteristics and propensity scores, the hazard ratios (HRs) of starting a lona, Barcelona, Spain
first glucose-lowering medication were 0.78 (95% CI 0.62–0.98) for Med-EatPlan + 13
Instituto Universitario de Investigaciones Bio-
EVOO and 0.89 (0.71–1.12) for Med-EatPlan + nuts, compared with the control médicas y Sanitarias (IUIBS), Universidad de Las
Palmas de Gran Canaria, and Complejo Hospi-
eating plan. After a median follow-up of 5.1 years, the adjusted HRs of starting
talario Universitario Insular Materno Infantil
insulin treatment were 0.87 (0.68–1.11) for Med-EatPlan + EVOO and 0.89 (0.69– (CHUIMI), Servicio Canario de Salud, Las Palmas
1.14) for Med-EatPlan + nuts compared with the control eating plan. de Gran Canaria, Spain
14
Lipid Unit, Department of Internal Medicine,
CONCLUSIONS Bellvitge Biomedical Research Institute-Hospital
Universitari de Bellvitge, L’Hospitalet de Llobre-
Among participants with type 2 diabetes, a Med-EatPlan + EVOO may delay the
gat, Barcelona, Spain
introduction of new-onset glucose-lowering medications. The Med-EatPlan did not 15
Department of Cardiology, University Hospital
result in a significantly lower need for insulin. Araba, Vitoria, Spain
care.diabetesjournals.org Basterra-Gortari and Associates 1391

Diabetes has reached epidemic propor- on the need for a first glucose-lowering concealment was achieved by using
tions, and this disease is at the forefront of medication (either oral or injectable) com- closed envelopes during part of the pilot
public health problems, affecting 451 mil- pared with a low-fat (control) eating plan phase of the study, but envelopes were
lion people worldwide in 2017 (1). More among trial participants with type 2 diabetes not used for the rest of the study. A
than 90% of patients with diabetes have who did not require glucose-lowering med- computer-generated random number se-
type 2 diabetes (2). The attainment and ication at enrollment. In addition, we sep- quence provided randomization tables for
maintenance of good glycemic control arately assessed the initiation of insulin 11 study sites, which included 169 clinics.
reduces the risk of long-term complica- treatment as a second outcome. These tables had four strata (women ,70
tions of type 2 diabetes (3). However, years of age, women $70 years of age,
glucose levels increase over the natural RESEARCH DESIGN AND METHODS men ,70 years of age, and men $70
history of type 2 diabetes (4,5), and this The PREDIMED study was designed as a years of age). In a subset of participants
progressive nature of the disease usually parallel-group,multicenter,randomizedtrial. (15% of the participants with type 2 di-
requires the sequential addition of glu- It was conducted in Spain to assess the abetes), there were deviations from the
cose-lowering medications (5). effects of two Med-EatPlans versus a low-fat randomization procedures as reported in
A healthful eating pattern, such as the control eating plan on the primary preven- detail elsewhere (15). To summarize, par-
Mediterranean eating plan (Med-EatPlan), tion of cardiovascular disease in adults at ticipants who lived in the same household
is a key component of type 2 diabetes high risk but without previously documented of previously randomized participants
management (6,7). The traditional cardiovascular disease at baseline. Detailed (usually their spouses) were assigned
Mediterranean pattern is characterized methods of the trial have been published to the same intervention (since enroll-
by a high intake of olive oil, fruits, veg- previously (14,15) and are available at ment) as their spouses already in the
etables, nuts, and cereals; a moderate www.predimed.es. trial. In addition, a subgroup of 311
intake of fish and poultry; a low intake of The trial was conducted in 11 recruiting participants of 1 of the 11 participating sites
red meat, whole-fat diary, and sweet centers affiliated with 11 Spanish uni- (site D) were not individually randomized
desserts; and wine consumption with versity hospitals. A total of 7,447 partic- but, instead, were assigned in small clus-
meals is allowed in moderation (8). Well- ipants underwent randomization from ters according to the clinic where they
conducted and analyzed prospective co- October 2003 through June 2009. Eligible belonged (i.e., all adults in the same clinic
horts (9,10) have consistently supported participants were men (55–80 years of received the same intervention).
the effectiveness of the Med-EatPlan for age) and women (60–80 years of age) ParticipantsassignedtotheMed-EatPlan+
reducing the incidence of type 2 diabetes, free of cardiovascular disease at enroll- EVOO received 1 L of EVOO per week
and a large intervention study, the Pre- ment who had either type 2 diabetes or at for free, and they were recommended to
vención con Dieta Mediterránea (PRE- least three of the following major car- meet the goal of consuming at least 4
DIMED) trial, showed that a Med-EatPlan diovascular risk factors: current smoking, tablespoons/day. Participants allocated to
supplemented with either extra-virgin olive hypertension, elevated LDL cholesterol the Med-EatPlan + nuts received 30 g/day
oil (EVOO) or mixed nuts was superior to a levels, low HDL cholesterol levels, over- of mixed nuts (15 g walnuts, 7.5 g hazelnuts,
low-fat diet for the prevention of type 2 weight or obesity, or a family history of and 7.5 g almonds), also at no cost. Par-
diabetes (11,12). Previously, a trial con- premature coronary heart disease. Detailed ticipants in the control group received
ducted in patients with newly diagnosed enrollment criteria have been published small nonfood gifts. Neither energy re-
type 2 diabetes found that compared previously (14,15). The protocol was ap- striction nor increased physical activity was
with a low-fat diet, an energy-restricted proved by the institutional review boards at promoted for any of the study groups.
Med-EatPlan allows for better glycemic all study locations. All participants pro- A general medical questionnaire,
control and delays the need for new-onset vided written informed consent. a 137-item validated food frequency
glucose-lowering medications (13). How- The protocol specified that partici- questionnaire (16), and the validated
ever, the potential preventive role of the pants were to be randomized in a 1:1:1 Spanish version of the Minnesota Lei-
Med-EatPlan for delaying the progression of ratio to one of three dietary interven- sure Time Physical Activity Questionnaire
type 2 diabetes, without energy restriction, tions: a Med-EatPlan supplemented with were administered at randomization
weight loss, or other lifestyle interventions, EVOO (Med-EatPlan + EVOO), a Med- and yearly thereafter (14). Information
has not been assessed in a clinical trial. EatPlan supplemented with mixed nuts from the food frequency questionnaire
In this subgroup analysis of the (Med-EatPlan + nuts), or a control eating was used to calculate energy and nutri-
PREDIMED trial, we tested the effect plan that consisted of advice to reduce ent intake. Weight, height, and waist cir-
of the two supplemented Med-EatPlans intake of all types of fat. Allocation cumference were directly measured (17).

16
Department of Endocrinology and Nutrition, Received 2 December 2018 and accepted 14 May © 2019 by the American Diabetes Association.
Complejo Hospitalario de Navarra, Instituto de 2019 Readers may use this article as long as the
investigación Sanitaria de Navarra (IdiSNA), This article contains Supplementary Data online at work is properly cited, the use is educational
Pamplona, Spain http://care.diabetesjournals.org/lookup/suppl/ and not for profit, and the work is not altered.
17
University of the Basque Country UPV/EHU, doi:10.2337/dc18-2475/-/DC1. More information is available at http://www
Vitoria-Gasteiz, Spain .diabetesjournals.org/content/license.
This article is featured in a podcast available at
Corresponding author: Fernando Arós, lfaborau@ http://www.diabetesjournals.org/content/diabetes-
gmail.com core-update-podcasts.
Clinical trial reg. no. ISRCTN35739639, www.isrctn.org
1392 Mediterranean Eating Plan and Diabetes Therapy Diabetes Care Volume 42, August 2019

Figure 1—Study flowchart.

For participants in the two Med- was used to assess adherence to the included data only on participants with
EatPlan groups, dietitians ran individual control eating plan. During follow-up, type 2 diabetes and not using insulin at
and group dietary training sessions at scores on the 14-item Med-EatPlan ques- baseline (n = 3,230). Among these 3,230
the baseline visit and quarterly there- tionnaire increased for the participants participants, 2,020 were receiving at least
after. In each session, a validated 14- randomized to the two Med-EatPlan one oral agent at baseline and were
item dietary questionnaire was used to groups (15,19). Biomarkers also showed excluded in the analyses of new-onset
estimate adherence to either of the that the intervention changed the overall glucose-lowering medications (Fig. 1).
Med-EatPlans (18). The answers to these dietary pattern of participants. Specif- In the time-to-event analyses, we as-
questionnaires were used as a tool to ically, adherence to the Med-EatPlan + sessed two outcomes: 1) introduction
tailor the intervention for each partici- EVOO intervention was examined by of the first glucose-lowering medica-
pant and to negotiate changes to up- measuring urinary hydroxytyrosol (a tion (oral or injectable) among partic-
grade participants’ adherence. Participants biomarker of EVOO consumption), and ipants on only lifestyle management
in the control group also received adherence to the Med-EatPlan + nuts at enrollment and 2) insulin initiation.
dietary training at the baseline visit intervention was examined by measur- During the trial, participants’ physi-
and completed the 14-item dietary ques- ing the plasma proportion of a-linolenic cians adjusted glucose-lowering med-
tionnaire used to examine baseline acid (a fatty acid characteristic of ications at their discretion to achieve
adherence to Med-EatPlan. Through Oc- walnuts). The blood and urine samples individually appropriate glycemic tar-
tober 2006, participants in the control were taken at 1, 3, and 5 years of gets. Glucose-lowering medications
group received only a leaflet describing follow-up in random subsamples of par- were obtained from the questionnaires
the low-fat eating plan. Thereafter, par- ticipants (15). completed by the participants at base-
ticipants assigned to the control eating Among the initial 7,447 participants of line and yearly thereafter. Nurses and
plan also received personalized advice the total PREDIMED trial, we excluded research assistants who collected this
and were invited to group sessions with those without diabetes at baseline (n = information were blinded with respect to
the same frequency and intensity as 3,833). We also excluded participants the hypotheses of the current study. Other
those in the Med-EatPlan groups. A who received insulin at enrollment investigators assessing the outcomes
separate nine-item dietary questionnaire (n = 384). Finally, the current study were also blinded to these hypotheses.
care.diabetesjournals.org Basterra-Gortari and Associates 1393

Table 1—Baseline characteristics of participants according to intervention arm

Med-EatPlan + EVOO Med-EatPlan + nuts Control eating plan
Variable (n = 1,158) (n = 1,017) (n = 1,055)
Age (years), mean (SD) 67.5 (6.2) 67.1 (6.1) 67.7 (6.5)
Female sex, n (%) 635 (54.8) 481 (47.3) 562 (53.3)
BMI (kg/m2)
Mean (SD) 29.7 (3.8) 29.7 (3.9) 30.2 (4.3)
,25, n (%) 116 (10.0) 105 (10.3) 92 (8.7)
25–30, n (%) 519 (44.8) 448 (44.1) 454 (43.0)
.30, n (%) 523 (45.2) 464 (45.6) 509 (48.3)
Body weight (kg), mean (SD) 76.3 (11.8) 77.1 (12.0) 77.2 (12.7)
Married, n (%) 921 (79.5) 783 (77.0) 790 (74.9)
Smoking status, n (%)
Never 714 (61.7) 581 (57.1) 646 (61.2)
Former 301 (26.0) 308 (30.3) 280 (26.5)
Current 143 (12.4) 128 (12.6) 129 (12.2)
Waist circumference (cm), mean (SD) 101 (10) 101 (10) 102 (11)
Waist-to-height ratio, mean (SD) 0.63 (0.06) 0.63 (0.06) 0.64 (0.07)
Hypertension, n (%) 847 (73.1) 722 (71.0) 793 (75.2)
Dyslipidemia, n (%) 685 (59.2) 600 (59.0) 621 (58.9)
Medication use, n (%)
Oral glucose-lowering medications 711 (61.4) 623 (61.3) 686 (65.0)
Lipid-lowering drugs 545 (47.1) 456 (44.8) 495 (46.9)
Antihypertensive agents 774 (66.8) 651 (64.0) 708 (67.1)
Leisure time physical activity level (MET min/day), mean (SD) 233 (236) 257 (258) 226 (261)
Data are mean 6 SD or n (%). BMI is weight in kilograms divided by the square of height in meters. The waist-to-height ratio is waist circumference
divided by height. Hypertension was defined as a systolic blood pressure of $140 mmHg, a diastolic blood pressure of $90 mmHg, or the use of
antihypertensive therapy. Dyslipidemia was defined as an LDL cholesterol level .160 mg/dL (4.1 mmol/L), an HDL cholesterol level of #40 mg/dL
(1.0 mmol/L) in men or #50 mg/dL (1.3 mmol/L) in women, or the use of lipid-lowering therapy.

Statistical Analysis intervention groups. The model was also estimate the probability of requiring in-
All analyses were performed on an inten- adjusted for hypertension (yes/no), sulin therapy during follow-up.
tion-to-treat basis. We assessed the effect dyslipidemia (yes/no), smoking status All P values are two-tailed at the ,0.05
of the intervention on the need for glucose- (never smoked, former smoker, or cur- level. We used Stata version 12.0 statis-
lowering medications fitting Cox propor- rent smoker), BMI (continuous), waist- tical software.
tional hazard regression models. Hazard to-height ratio (continuous), leisure
ratios (HRs) and their 95% CIs were calcu- time physical activity (continuous), and RESULTS
lated, considering the control group as the total energy intake (continuous). For
reference. Person-years of follow-up were We assessed 1,158, 1,017, and 1,055 par-
the assessment of the second outcome,
calculated from baseline to the earliest ticipants from the Med-EatPlan + EVOO,
namely insulin initiation, the models
event (glucose-lowering medication), loss the Med-EatPlan + nuts, and the control
were also adjusted for baseline oral
to follow-up, or end of follow-up (December agents (yes/no). Robust variance estima- eating plan, respectively. These 3,230 par-
1, 2010). We repeated the analyses using tors were used to account for intracluster ticipants had type 2 diabetes and were not
insulin initiation as the dependent variable. correlation in Cox models, considering as treated with insulin at enrollment. Baseline
To address the small departures from clusters the members of the same house- characteristics were well balanced in the
individual randomization in a subset of hold and the participants in the same three study groups without any clinically
participants, we conducted analyses that clinic of site D allocated in clusters. As a significant between-group differences
did not assume that all the participants sensitivity analysis, we removed partic- (Table 1). Perhaps the only exception was
were randomly allocated and that ran- ipants whose randomization procedures the lower proportion of women (absolute
domization would distribute baseline had deviated from protocol: second difference 6%) in the Med-EatPlan + nuts
characteristics of the participants equally members of the same household and group compared with the control group.
across intervention groups. Thus, in ad- all participants from site D. We repeated In any case, we always adjusted for sex.
dition to the crude model, in a subse- all analyses after merging the two Med- During follow-up, the mean scores on
quent multivariable model, we stratified EatPlan groups and assessed their ef- the 14-item Med-EatPlan questionnaire
by sex, age (deciles), recruiting center, fect compared with the control group. increased in both Med-EatPlan groups
and educational level (five categories) We used the Kaplan-Meier method to and were higher than in the control group
and adjusted for propensity scores that describe the probability of remaining (Supplementary Fig. 1). Supplementary
used 30 baseline variables to estimate the free of glucose-lowering medications Table 1 shows the mean nutrient changes
probability of assignment to each of the and Nelson-Aalen incidence curves to in the three groups.
1394 Mediterranean Eating Plan and Diabetes Therapy Diabetes Care Volume 42, August 2019

eating plan. When we assessed the two

Med-EatPlan groups together, the HR
of starting glucose-lowering medica-
tion was 0.91 (0.74–1.11). The propen-
sity score and multivariable-adjusted
HRs of starting long-term insulin treat-
ment were 0.87 (0.68–1.11) for Med-
EatPlan + EVOO and 0.89 (0.69–1.14)
for Med-EatPlan + nuts, using the con-
trol eating plan as the reference. The
adjusted HR for the Med-EatPlan groups
(both groups merged vs. the control
group) was 0.88 (0.71–1.09). After ex-
cluding second members of the same
household and all participants from site D
(165 and 311, respectively), the analy-
sis with 2,754 individuals showed an
adjusted HR, for both Med-EatPlan
combined versus the control eating
plan group, of 0.92 (0.73–1.16). After
1-year follow-up, a 1-unit increase in
the score on the 14-item Med-EatPlan
questionnaire was associated thereafter
with an adjusted HR of starting insulin of
0.95 (0.88–1.01). The mean fasting blood
Figure 2—Kaplan-Meier estimate of the probability of remaining free of glucose-lowering medications. glucose level was 145 6 40 mg/dL at
*The Cox model was stratified according to sex, age (deciles), recruiting center, and educational baseline and 143 6 42 mg/dL after 5 years
level (five categories) and adjusted for propensity scores that used 30 baseline variables to
estimate the probability of assignment to each of the intervention groups. The model was also
in the Med-EatPlan + EVOO group, 144 6
adjusted for hypertension (yes/no), dyslipidemia (yes/no), smoking status (never smoked, former 42 mg/dL at baseline and 140 6 37 mg/dL
smoker, or current smoker), BMI (continuous), waist-to-height ratio (continuous), leisure time after 5 years in the Med-EatPlan + nuts,
physical activity (continuous), and total energy intake (continuous). Robust SEs to account for and 147 6 43 mg/dL at baseline and
intracluster correlations were used. Med, Med-EatPlan. 146 6 46 mg/dL after 5 years in the
control group.

After a median follow-up of 3.2 years, 0.98) for Med-EatPlan + EVOO and 0.89
686 participants with only lifestyle man- (0.71–1.12) for Med-EatPlan + nuts com- CONCLUSIONS
agement at baseline started glucose- pared with the control eating plan. When In this trial, a Med-EatPlan supple-
lowering medications (576 participants both Med-EatPlan groups were merged mented with EVOO without any caloric
started an oral agent, 37 participants together, we found an HR of 0.83 (0.68– restriction or weight-loss goals, but not
started long-term insulin, and 73 partic- 1.02). In a sensitivity analysis, when we a Med-EatPlan supplemented with nuts,
ipants started both an oral agent and excluded second members of the same significantly decreased the need of
insulin at the same time). After a median household (56 participants) and all par- new-onset pharmacologic interventions,
follow-up of 5.1 years, a total of 407 ticipants from site D (141 participants), compared with a control eating plan,
insulin-naı̈ve participants at baseline the results with 1,013 individuals aligned in participants with type 2 diabetes
started long-term insulin therapy. with the findings of the adjusted model. and no cardiovascular disease at en-
Figure 2 shows the probability of The adjusted HR for both Med-EatPlan rollment after a median follow-up of
remaining free of glucose-lowering groups merged together was 0.85 (0.69– 3.2 years. A Med-EatPlan + EVOO or
medications in the three groups. The 1.05). After 1-year follow-up, a 1-unit nuts did not result in a lower rate of
unadjusted HRs of starting glucose- increase in the score on the 14-item Med- insulin initiation after a median follow-up
lowering medications were 0.83 (95% EatPlan questionnaire was associated of 5.1 years.
CI 0.69–0.99) for a Med-EatPlan + thereafter with an adjusted HR of starting The lower need of starting a first
EVOO and 0.92 (0.76–1.11) for a Med- glucose-lowering medication of 0.98 glucose-lowering medication (either
EatPlan + nuts compared with the con- (0.92–1.05). oral or injectable) with the Med-EatPlan +
trol eating plan. When we assessed the two Figure 3 shows the probability of re- EVOO probably reflects the better gly-
Med-EatPlan groups together, the HR maining free of insulin in the three cemic control of this group during the
of starting glucose-lowering medication groups. The unadjusted HRs of starting long follow-up of the PREDIMED study,
was 0.87 (0.74–1.02). The multivariable- long-term insulin treatment were 0.90 and for this reason, a first treatment was
adjusted HRs, including adjustments for (95% CI 0.72–1.14) for Med-EatPlan + prescribed less often to achieve or main-
propensity scores, of starting glucose- EVOO and 0.91 (0.71–1.16) for Med- tain glycemic goals. The favorable effect
lowering medications were 0.78 (0.62– EatPlan + nuts compared with the control was likely due to the overall composition
care.diabetesjournals.org Basterra-Gortari and Associates 1395

(25) randomly assigned 259 patients with

type 2 diabetes to one of three diets: low-
carbohydrate Mediterranean, traditional
Mediterranean, and the 2003 American
Diabetes Association diet. The mean
weight loss for the three diets was
10.1, 7.4, and 7.7 kg, respectively. Using
as a reference the American Diabetes
Association diet, Elhayany et al. reported
greater reductions in HbA1c levels in
participants allocated to the low-
carbohydrate Mediterranean diet and
the traditional Mediterranean diet (av-
erage difference changes of 0.4% and
0.2%, respectively). In a subset of the
PREDIMED trial, better adherence to the
Med-EatPlan was associated with lower
HbA 1c levels, although the observed
differences were statistically nonsignif-
icant (26). These previous results pro-
vide support to the benefits of the
Med-EatPlan + EVOO that we have
observed.
Med-EatPlan + nuts was also associ-
ated with a lower need of antihypergly-
Figure 3—Nelson-Aalen estimate of the probability of requiring insulin therapy. *The Cox model cemic drug therapy in the point estimate,
was stratified according to sex, age (deciles), recruiting center, and educational level (five but the CIs were wider, and the upper
categories) and adjusted for propensity scores that used 30 baseline variables to estimate the
limit was compatible with a 12% higher
probability of assignment to each of the intervention groups. The model was also adjusted for
hypertension (yes/no), dyslipidemia (yes/no), smoking status (never smoked, former smoker, risk. This finding contrasts with that in the
or current smoker), BMI (continuous), waist-to-height ratio (continuous), leisure-time physical Med-EatPlan + EVOO group. The differ-
activity (continuous), total energy intake (continuous), and oral agents (yes/no). Robust SEs to ence in the effects of the two interven-
account for intracluster correlations were used. Med, Med-EatPlan. tions using the same Med-EatPlan as
the background diet might be related
to several factors. It is possible that there
of the dietary pattern and not to de- studies (122,810 subjects) and one ran- are differences between EVOO and
creased caloric intake, increased physical domized controlled trial (PREDIMED), nuts. A meta-analysis in patients with
activity, or weight loss because such life- greater adherence to a Med-EatPlan was type 2 diabetes reported that EVOO
style interventions were not part of the associated with a significant 19% lower supplementation resulted in a change
PREDIMED trial, and there were no nota- risk of type 2 diabetes (9). In agree- in HbA 1c of 20.27% (95% CI 20.37
ble between-group differences in these ment with these results, the initial to 20.17%) (27). Nuts have been asso-
characteristics at baseline or during fol- 3-month assessment in 772 participants of ciated with a lower risk of type 2 diabetes
low-up (20). In particular, after adjust- the PREDIMED study found an improved (28). However, the glycemic effect of nut-
ment for propensity scores and use of fasting glucose in the Med-EatPlan groups enriched meals may be lower in people
robust variance estimators, the average in the absence of weight loss (22). In with diabetes than in people without
difference in body weight change at addition, two randomized trials also re- diabetes (29). In addition, at the end
5 years in the Med-EatPlan + EVOO group ported an improvement in glycemic con- of PREDIMED, 22% of total calories in the
was 20.41 kg (95% CI 20.83 to 0.01 kg), trol of the Med-EatPlan combined with Med-EatPlan + EVOO group were from
and in the Med-EatPlan + nuts group, it other lifestyle strategies, such as exercise EVOO, whereas only 8% of calories in the
was 20.02 kg (20.45 to 0.42 kg) com- or calorie-restricted diets (23,24). In a Med-EatPlan + nuts group were from
pared with the control group (20). In 4-year trial (the longest to date), Esposito nuts. However, the CIs for both estimates
addition, no between-group difference et al. (13) randomized 215 patients with were widely overlapping.
in body weight was found in participants newly diagnosed type 2 diabetes to a low- Our results suggest a 12% lower rate of
with baseline diabetes (20). carbohydrate Mediterranean-style diet initiation of insulin in the point estimate.
Previously, the PREDIMED trial re- or a low-fat diet. At the end of the trial, Nonetheless, a 30% lower risk and a 10%
ported a significant reduction in the 44% of patients in the Mediterranean- higher risk are also reasonably compat-
risk of type 2 diabetes among partic- style diet group and 70% in the ible with our data. This highlights possi-
ipants without diabetes at baseline low-fat group required glucose-lowering ble differences among participants of
(11,12,21). In a meta-analysis of prospec- medications. Participants randomized PREDIMED because participants who ini-
tive studies published between 2007 and to the Med-EatPlan lost more weight. tiated insulin therapy usually had a lon-
2014, including eight prospective cohort Finally, in a 12-month trial, Elhayany et al. ger duration of diabetes and a higher
1396 Mediterranean Eating Plan and Diabetes Therapy Diabetes Care Volume 42, August 2019

HbA1c than those on lifestyle manage- Funding. The supplemental foods used in the Author Contributions. F.J.B.-G., M.R.-C.,
ment. Differences between participants study were donated by Patrimonio Comunal M.A.M.-G., N.B., J.V.S., M.Fit., E.R., E.G.-G.,
Olivarero and Hojiblanca, Madrid, Spain (EVOO); M.Fio., J.L., R.E., L.S.-M., X.P., J.I.G., M.B., O.C.,
who initiated insulin and those in- the California Walnut Commission, Sacramento, CA Á.A.-G., L.F., and F.A. revised the manuscript for
cluded in diabetes prevention analyses (walnuts); and Borges SA (almonds) and La Morella important intellectual content and read and ap-
of PREDIMED are even greater (11,12). Nuts (hazelnuts), Reus, Spain. The PREDIMED trial proved the final manuscript. F.J.B.-G., M.R.-C.,
However, other lifestyle interventions was supported by Instituto de Salud Carlos III, the M.A.M.-G., M.Fit., E.R., E.G.-G., M.Fio., J.L., R.E.,
have shown a lower need of insulin official funding agency for biomedical research of L.S.-M., X.P., L.F., and F.A. acquired, analyzed, or
the Spanish government, through grants provided interpreted data. F.J.B.-G., M.A.M.-G., and F.A.
in participants with diabetes. Partici- to research networks specifically developed for the drafted the manuscript. M.A.M.-G., M.Fit., E.R.,
pants randomized to intensive lifestyle trial (RTIC RD 06/0045 [coordinator: M.A.M.-G.] and E.G.-G., M.Fio., R.E., L.S.-M., and F.A. conceived
intervention, focusing on weight loss, in RTIC G03/140 [coordinator: R.E.]). All investigators the study concept and design. M.A.M.-G., E.R.,
the Look AHEAD (Action for Health in of the PREDIMED trial belong to CIBER, an initiative J.L., R.E., and L.S.-M. obtained funding. F.J.B.-G.,
of Instituto de Salud Carlos III. The authors also M.A.M.-G., and F.A. are the guarantors of this
Diabetes) trial had a lower use of insulin
acknowledge grants from the National Institutes work and, as such, had full access to all of the data
than participants in the control group (30). of Health Clinical Center (1R01-HL-118264-01 in the study and take responsibility for the
Our study has certain limitations. First, and 1R01-DK-102896), Fondo de Investigación integrity of the data and the accuracy of the
the need for glucose-lowering medica- Sanitaria–Fondo Europeo de Desarrollo Regional data analysis.
tions was not a prespecified end point in (PI04/0233, PI05/0976, PI07/0240, PI10/01407,
the PREDIMED trial. Thus, these analyses PI10/02658, PI11/00049, PI11/02505, and References
AGL2010-22319-C03-03), Consejerı́a de Salud Junta 1. Cho NH, Shaw JE, Karuranga S, et al. IDF
are exploratory. In addition, the analyses de Andalucı́a (PI0105/2007), and the Generalitat Diabetes Atlas: global estimates of diabetes
of this study were conducted in the Valenciana, Valencia, Spain (ACOMP/2013/165 prevalence for 2017 and projections for 2045.
subgroup of participants with type 2 di- and COMP/2013/159). Diabetes Res Clin Pract 2018;138:271–281
abetes. However, there is no reason to Duality of Interest. E.R. reports grants, non- 2. Xu G, Liu B, Sun Y, et al. Prevalence of
suspect that the randomization would financial support, and other fees from the Cal- diagnosed type 1 and type 2 diabetes among
ifornia Walnut Commission; grants, personal US adults in 2016 and 2017: population based
not have worked in such a large number fees, nonfinancial support, and other fees from study. BMJ 2018;362:k1497
of participants. Second, we recruited Merck Sharp & Dohme, Alexion, and Ferrer 3. UK Prospective Diabetes Study (UKPDS)
white adults (55–80 years of age) without International; personal fees, nonfinancial sup- Group. Intensive blood-glucose control with
previously documented cardiovascular port, and other fees from Aegerion, Amarin, and sulphonylureas or insulin compared with con-
disease at baseline. Thus, the results Danone; grants and personal fees from Sanofi; ventional treatment and risk of complications in
and grants from Amgen and Pfizer outside of the patients with type 2 diabetes (UKPDS 33). Lancet
cannot be generalized to all subjects submitted work. R.E. reports serving on the 1998;352:837–853
with type 2 diabetes. Third, inherent board of and receiving lecture fees from 4. Kahn SE, Haffner SM, Heise MA, et al.;
to the design of a dietary intervention the Research Foundation on Wine and Nutri- ADOPT Study Group. Glycemic durability of
trial using a whole dietary pattern, the tion; serving on the boards of the Beer and rosiglitazone, metformin, or glyburide mono-
trial could not be double blind. In any Health Foundation and the European Founda- therapy. N Engl J Med 2006;355:2427–2443
tion for Alcohol Research; receiving lecture fees 5. Turner RC, Cull CA, Frighi V, Holman RR; UK
case, participants and staff members from Instituto Cerventes, Fundación Dieta Prospective Diabetes Study (UKPDS) Group. Glyce-
involved in the intervention and data Mediterránea, Cerveceros de Espa~ na, Lilly Lab- mic control with diet, sulfonylurea, metformin, or
collection were unaware of the hypoth- oratories, AstraZeneca, and Sanofi; receiving insulin in patients with type 2 diabetes mellitus:
eses of the present report. The strengths consultancy fees from KAO Coorporation; and progressive requirement for multiple therapies
of the PREDIMED trial include the large receiving grant support through his institution (UKPDS 49). JAMA 1999;281:2005–2012
from Novartis, Amgen, Bicentury, and Grand 6. Evert AB, Boucher JL, Cypress M, et al. Nu-
sample size, long follow-up period, Fountaine. L.S.-M.reportsserving onthe boardsof trition therapy recommendations for the man-
breadth of included participants with the Mediterranean Diet Foundation and the Beer agement of adults with diabetes. Diabetes Care
type 2 diabetes, and adjustment for a and Health Foundation. X.P. reports serving on 2014;37(Suppl. 1):S120–S143
wide array of potential confounders in the board of and receiving grant support through 7. Davies MJ, D’Alessio DA, Fradkin J, et al.
multivariable analyses. his institution from the Residual Risk Reduction Management of hyperglycemia in type 2 diabe-
Initiative Foundation; serving on the board of tes, 2018. A consensus report by the American
In summary, our study results show Omegafort; serving on the board of and receiv- Diabetes Association (ADA) and the European
that PREDIMED participants with type 2 ing payment for the development of educational Association for the Study of Diabetes (EASD).
diabetes who underwent an interven- presentations as well as grant support through Diabetes Care 2018;41:2669–2701
tion with an energy-unrestricted Med- his institution from Ferrer International; receiv- 8. Willett WC, Sacks F, Trichopoulou A, et al.
EatPlan + EVOO had significantly lower rates ing consulting fees from Abbott Laboratories; Mediterranean diet pyramid: a cultural model
receiving lecture fees as well as grant support for healthy eating. Am J Clin Nutr 1995;61(Suppl.):
of initiation of glucose-lowering medica- through his institution from Merck and Roche; 1402S–1406S
tions. Our results are compatible not only receiving lecture fees from Danone and Esteve; 9. Schwingshackl L, Missbach B, König J, Hoffmann
with a benefit of a Med-EatPlan + nuts receiving payment for the development of ed- G. Adherence to a Mediterranean diet and risk of
in the rates of initiation of glucose- ucational presentations from Menarini, Mylan, diabetes: a systematic review and meta-analysis.
lowering medications and with a benefit LACER, and Rubio Laboratories; and receiving Public Health Nutr 2015;18:1292–1299
grant support through his institution from Sanofi, 10. Koloverou E, Esposito K, Giugliano D,
of a Med-EatPlan + EVOO or nuts in the need Kowa, Unilever, Boehringer Ingelheim, and Karo Panagiotakos D. The effect of Mediterranean
of insulin but also with a slightly higher risk. Bio. F.A. reports receiving grants from Instituto diet on the development of type 2 diabetes
de Salud Carlos III, Fondo de Investigación Sanitario mellitus: a meta-analysis of 10 prospective stud-
(FIS), and CIBEROBN during the conduct of the ies and 136,846 participants. Metabolism 2014;
study and personal fees from payment for the 63:903–911
Acknowledgments. The authors thank the development of educational presentations out- 11. Salas-Salvadó J, Bulló M, Babio N, et al.;
participants for involvement in the trial, the side the submitted work. No other potential PREDIMED Study Investigators. Reduction in the
PREDIMED personnel, and the personnel of all conflicts of interest relevant to this article were incidence of type 2 diabetes with the Mediter-
affiliated primary care centers. reported. ranean diet: results of the PREDIMED-Reus
care.diabetesjournals.org Basterra-Gortari and Associates 1397

nutrition intervention randomized trial. Diabetes adherence among older Spanish men and women. 25. Elhayany A, Lustman A, Abel R, Attal-Singer J,
Care 2011;34:14–19 J Nutr 2011;141:1140–1145 Vinker S. A low carbohydrate Mediterranean diet
12. Salas-Salvadó J, Bulló M, Estruch R, et al. 19. Zazpe I, Sanchez-Tainta A, Estruch R, et al. A improves cardiovascular risk factors and diabetes
Prevention of diabetes with Mediterranean di- large randomized individual and group interven- control among overweight patients with type 2
ets: a subgroup analysis of a randomized trial. tion conducted by registered dietitians increased diabetes mellitus: a 1-year prospective random-
Ann Intern Med 2014;160:1–10 adherence to Mediterranean-type diets: the ized intervention study. Diabetes Obes Metab
13. Esposito K, Maiorino MI, Ciotola M, et al. PREDIMED study. J Am Diet Assoc 2008;108: 2010;12:204–209
Effects of a Mediterranean-style diet on the 1134–1144; discussion 1145 26. Dı́ez-Espino J, Buil-Cosiales P, Serrano-
need for antihyperglycemic drug therapy in 20. Estruch R, Martı́nez-González MA, Corella D, Martı́nez M, Toledo E, Salas-Salvad ó J,
patients with newly diagnosed type 2 diabe- et al.; PREDIMED Study Investigators. Effect of a Martı́nez-González MÁ. Adherence to the Med-
tes: a randomized trial. Ann Intern Med 2009; high-fat Mediterranean diet on bodyweight and iterranean diet in patients with type 2 diabetes
151:306–314 waist circumference: a prespecified secondary mellitus and HbA1c level. Ann Nutr Metab 2011;
14. Martı́nez-González MÁ, Corella D, Salas- outcomes analysis of the PREDIMED randomised 58:74–78
Salvadó J, et al.; PREDIMED Study Investigators. controlled trial. Lancet Diabetes Endocrinol 27. Schwingshackl L, Lampousi AM, Portillo MP,
Cohort profile: design and methods of the 2019;7:e6–e17 Romaguera D, Hoffmann G, Boeing H. Olive oil in
PREDIMED study. Int J Epidemiol 2012;41:377–385 21. Martı́nez-González MA, Salas-Salvadó J, the prevention and management of type 2 di-
15. Estruch R, Ros E, Salas-Salvadó J, et al.; Estruch R, Corella D, Fitó M, Ros E; PREDIMED abetes mellitus: a systematic review and meta-
PREDIMED Study Investigators. Primary preven- Investigators. Benefits of the Mediterranean analysis of cohort studies and intervention trials.
tion of cardiovascular disease with a Mediterra- diet: insights from the PREDIMED study. Prog Nutr Diabetes 2017;7:e262
nean diet supplemented with extra-virgin olive oil Cardiovasc Dis 2015;58:50–60 28. Jiang R, Manson JE, Stampfer MJ, Liu S,
or nuts. N Engl J Med 2018;378:e34 22. Estruch R, Martı́nez-González MA, Corella D, Willett WC, Hu FB. Nut and peanut butter
16. Fernández-Ballart JD, Pi~nol JL, Zazpe I, et al. et al.; PREDIMED Study Investigators. Effects of consumption and risk of type 2 diabetes in
Relative validity of a semi-quantitative food- a Mediterranean-style diet on cardiovascular risk women. JAMA 2002;288:2554–2560
frequency questionnaire in an elderly Mediter- factors: a randomized trial. Ann Intern Med 2006; 29. Kendall CW, Esfahani A, Josse AR, Augustin
ranean population of Spain. Br J Nutr 2010;103: 145:1–11 LS, Vidgen E, Jenkins DJ. The glycemic effect of
1808–1816 23. Toobert DJ, Glasgow RE, Strycker LA, et al. nut-enriched meals in healthy and diabetic sub-
17. Salas-Salvadó J, Fernández-Ballart J, Ros E, Biologic and quality-of-life outcomes from the jects. Nutr Metab Cardiovasc Dis 2011;21(Suppl. 1):
et al.; PREDIMED Study Investigators. Effect of a Mediterranean Lifestyle Program: a randomized S34–S39
Mediterranean diet supplemented with nuts on clinical trial. Diabetes Care 2003;26:2288–2293 30. Wing RR, Bolin P, Brancati FL, et al.; Look
metabolic syndrome status: one-year results of 24. Shai I, Schwarzfuchs D, Henkin Y, et al.; AHEAD Research Group. Cardiovascular effects
the PREDIMED randomized trial. Arch Intern Med Dietary Intervention Randomized Controlled of intensive lifestyle intervention in type 2 di-
2008;168:2449–2458 Trial (DIRECT) Group. Weight loss with a low- abetes [published correction appears in N Engl J
18. Schröder H, Fitó M, Estruch R, et al. A short carbohydrate, Mediterranean, or low-fat diet. Med 2014;370:1866]. N Engl J Med 2013;369:
screener is valid for assessing Mediterranean diet N Engl J Med 2008;359:229–241 145–154