PREDIMED

Published

Estruch R, et al. "Primary prevention of cardiovascular disease with a mediterranean diet". The New England Journal of Medicine. 2013. 368(14):1279-1290.
PubMed • Full text • PDF

Clinical Question

In patients at high risk for CV disease, does a Mediterranean diet enriched with extra-virgin olive oil or nuts reduce MI, CVA, or CV death compared to a low-fat diet?

Bottom Line

In the initial 2013 publication, a Mediterranean diet with extra-virgin olive oil or nuts was thought to reduce rates of MI, CVA, or CV death in those at high risk for CV disease. In 2018, this trial was retracted and replaced for abnormalities in randomization. The updated publication documented similar findings but the authors were unable to confirm adherence to randomization schemes given missing documentation.

Major Points

Epidemiologic trials from the 1960s suggested that a so-called "Mediterranean diet" rich in fresh fruits, vegetables, fish, olive oil, and wine, was associated with decreased rates of CV disease. Multiple studies have demonstrated a mortality benefit from a Mediterranean or Mediterranean-like diet after MI, including DART (1989) and the Cardioprotective Diet Study (1992). Most recently, the Lyon Diet Heart Study (1999) demonstrated that a Mediterranean diet reduced recurrent CV events by 50-70% among MI patients. The role of the Mediterranean diet in primary prevention of CV disease hadn't been well established.

The 2013 Prevención con Dieta Mediterránea (Prevention with Mediterranean Diet; PREDIMED) trial randomized 7,447 Spanish patients at high risk for CVD to one of three diets: 1) a Mediterranean diet supplemented with extra-virgin olive oil, 2) a Mediterranean diet supplemented with nuts, or 3) a control diet encouraging low-fat items. The trial was terminated early after a median follow-up of 4.8 years, and demonstrated that both Mediterranean diet groups reached a statistically significant reduction in the rate of the composite primary outcome of MI, CVA, or CV death. This corresponded to an absolute reduction in three CV events per 1000 patient-years, or a 30% relative risk reduction. However, of the individual components of the primary outcome, only rate of CVA was significantly reduced.

Of note, the Look AHEAD trial (published later in 2013) randomized 5,145 overweight or obese patients with type 2 diabetes to an intensive weight loss intervention through increased exercise and caloric restriction (with <30% of total calories from fat) or an intervention focusing on general support and education about the management of T2DM. The intensive group had an increased rate of weight loss though did not have any difference in the rate of CVD or its complications at ~10 years of follow-up. While this trial did not directly address the role of a Mediterranean diet in T2DM, this diet has emerged as a potential therapy for glycemic control in diabetics. A 2014 trial by Esposito found that newly-diagnosed diabetic patients assigned to a Mediterranean diet had a longer duration of time until initiation of hypoglycemic medications than those assigned to a control diet.

In 2018 the original PREDIMED publication was retracted and replaced with a revised version for several reasons. First, there were >400 participants who were sharing a household with another participant already enrolled who were assigned non-randomly to the same group as their household member. This detail was known by the authors but not reported in the original manuscript. Second, one of the eleven centers (which itself had 11 clinics) stopped using the randomization scheme and instead randomized entire clinics to an intervention, not individual patients to an intervention. This affected >450 (6%) of participants. Third, another site that enrolled nearly 600 participants inconsistently used randomization tables. Finally, it was revealed that concealment of randomization was not used in this study. Concealment prevents the individual who is performing the randomization from knowing the group assignment of the next randomized person. It is considered a key component of RCTs and minimizes selection bias in studies. Inadequate concealment can be associated with exaggeration of the treatment effect, with one survey finding a 30% lower ("better") odds ratio with the intervention of interest among trials without appropriate concealment. In the supplementary appendix provided with the revised manuscript, the authors document inability to confirm appropriate randomization schemes at all sites given lack of documentation and the duration of time since the study was published. The authors performed several complicated analyses to attempt to control for these deficiences (6 models are presented on page 17 of the Supplementary Appendix) that all seem to confirm the original findings of this study. However, given the multiple deviations from the documented randomization scheme in the original protocol, the initial solid findings are now open to further scrutiny. Further, numerous secondary analyses using the PREDIMED database have been published in high-impact journals. The implication of these randomization abnormalities on the secondary analyses is unknown.

Guidelines

ADA (2013)

• Weight loss for all overweight or obese patients with T2DM (level A)
• Weight loss in the above population is likely effective at two years with a low-carbohydrate, low-fat calorie restricted, or Mediterranean diet (level A)

Design

• Multicenter, prospective, randomized, controlled trial
• N=7,447 patients at high risk for CVD
  • Mediterranean diet with olive oil (n=2,543)
  • Mediterranean diet with nuts (n=2,454)
  • Low-fat control diet (n=2,450)
• Setting: Multiple Spanish centers
• Enrollment: 2003-2010
• Median follow-up: 4.8 years
• Analysis: Intention-to-treat
• Primary outcome: Composite of MI, CVA, or CV death

Population

Inclusion Criteria

• Men 55-80 years or women 60-80 years in age without CVD
• DM or ≥3 of the following:
  • Smoker
  • HTN
  • LDL ≥160 mg/dL
  • HDL ≤40 mg/dL (HDL ≥60 is protective and cancels one other risk factor)
  • BMI ≥25 kg/m²
  • First degree relative with CVD before 55 years (male) or 65 years (female)

Exclusion Criteria

• Established CVD
• Expected survival <1 year
• HIV or other immunodeficiency
• Use of illicit drugs
• Alcoholism, alcohol abuse, or daily EtOH intake > 80 g
• BMI >40 kg/m²
• Medical condition precluding participation
• Difficulties or major inconvenience to change dietary habits
• Inability to follow Mediterranean diet (eg religious reasons, chewing abnormalities)
• Low likelihood of dietary change from Prochaska and DiClemente model
• Allergy or hypersensitivity to diet food
• Other investigational trials in the last year
• Institutionalization, lacking autonomy, unable to ambulate, no stable address, unable to come to the outpatient clinic every 3 months
• Illiteracy
• Acute infection until resolved for 3 months

Baseline Characteristics

From the Mediterranean diet with olive oil group.

• Demographics: Age 67 years, female 58.7%, white 97.1%
• Health data: BMI 30, waist circumference 100 cm, Waist:Height 0.63
• PMH: HTN 82.1%, T2DM 50.4%, HLD 71.6%
• Family history of premature CVD: 22.7%
• Smoking history:
  • Never: 61.8%
  • Former: 24.3%
  • Current: 13.9%
• Medications: ACE inhibitor 48.6%, diuretic 21.0%, other antihypertensive 28.5%, statin 40.9%, non-statin lipid medication 4.8%, insulin 4.9%, oral hypoglycemics 30.2%, antiplatelet 18.7%, HRT 2.8%

Interventions

• Patients were randomized to one of three groups:
  • Mediterranean diet with extra-virgin olive oil
  • Mediterranean diet with nuts
  • Low-fat control diet
• Mediterranean diet group participants had individual and group dietary training sessions four times yearly:
  • Dietitians insisted that at least two meals per day be eaten at home seated at a table for >20 minutes; they recommended the abundant use of olive oil, ≥2 daily servings of vegetables, ≥2 daily servings of fresh fruits, ≥3 weekly servings of legumes, ≥3 weekly servings of seafood, ≥1 weekly serving of nuts or seeds, poultry or rabbit rather than red meat, and to cook ≥2 times weekly with tomato, garlic, and onion.
  • Dietitians recommended against dairy, cold meat, pate/duck, sugared/carbonated beverages, pastries and fried foods, and sweets.
  • Wine was recommended as the main source of alcohol rather than beer or other spirits, with maximum 3 glasses of wine per day.
  • For the olive oil group, patients were given one liter per week of extra-virgin olive oil
  • For the nut group, patients received 2 kg walnuts, 1 kg almonds, 1kg hazelnuts every 3 months.
• Control diet group participants were given baseline dietary training only and received a pamphlet yearly until 2006 when they received the same frequency interventions as the Mediterranean groups:
  • Dietitians recommend the avoidance of all types of fat, recommending low meats and low-fat dairy, cereals, potatoes, pasta, rice, fruits, and vegetables.
  • Olive oil, nuts, and fatty meats (including fatty fish) were discouraged in the control group.

Outcomes

Comparisons are Mediterranean diet with olive oil vs. Mediterranean diet with nuts vs. low-fat control diet.

Primary Outcome

Composite of MI, CVA, or CV death

8.1 vs. 8.0 vs. 11.2 events per 1000 person-years

Mediterranean with olive oil vs. low fat: HR 0.69; 95% CI 0.53-0.91; P=0.008

Mediterranean with nuts vs. low fat: HR 0.72; 95% CI 0.54-0.97; P=0.03

Either Mediterranean vs. low fat: HR 0.70; 95% CI 0.55-0.89; P=0.003

Secondary Outcomes

CVA

4.1 vs. 3.1 vs. 5.9 events per 1000 person-years

Mediterranean with olive oil vs. low fat: HR 0.67; 95% CI 0.46-0.98; P=0.04

Mediterranean with nuts vs. low fat: HR 0.54; 95% CI 0.35-0.84; P=0.006

Either Mediterranean vs. low fat: HR 0.61; 95% CI 0.44-0.86; P=0.005

MI

3.1 vs. 3.0 vs. 3.9 events per 1000 person-years

Mediterranean with olive oil vs. low fat: HR 0.80; 95% CI 0.51-1.26; P=0.34

Mediterranean with nuts vs. low fat: HR 0.74; 95% CI 0.46-1.19; P=0.22

Either Mediterranean vs. low fat: HR 0.77; 95% CI 0.51-1.15; P=0.20

CV mortality

2.2 vs. 3.0 vs. 3.1 events per 1000 person-years

Mediterranean with olive oil vs. low fat: HR 0.69; 95% CI 0.41-1.16; P=0.17

Mediterranean with nuts vs. low fat: HR 1.01; 95% CI 0.61-1.66; P=0.98

Either Mediterranean vs. low fat: HR 0.83; 95% CI 0.54-1.29; P=0.41

All-cause mortality

10.0 vs. 11.2 vs. 11.7 events per 1000 person-years

Mediterranean with olive oil vs. low fat: HR 0.82; 95% CI 0.64-1.07; P=0.15

Mediterranean with nuts vs. low fat: HR 0.97; 95% CI 0.74-1.26; P=0.82

Either Mediterranean vs. low fat: HR 0.89; 95% CI 0.71-1.12; P=0.32

Self-reported adherence

8.7 vs. 8.7 vs. 8.4 points on a scale of 14 (higher indicates better adherence)

Subgroup Analysis

Among subgroups studied, there were no statistically significant differences in the rates of the primary outcome in either Mediterranean diet compared to the control diet. However, the P-value for interaction approached significance for BMI, suggesting that the diet may most benefit obese patients (P=0.05).

Criticisms

• The low fat control's protocol was changed during the trial and the lower intensity of the dietary intervention may have introduced bias
• Loss to follow up in the low fat group may have skewed results
• Participants lived in Mediterranean countries and were at high risk for CVD; thus, it's unclear if the results will be applicable to others
• A true low-fat diet was not achieved by the control group
• Unclear if olive oil or nuts would benefit other diets
• Adding olive oil or nuts has unclear effects on weight gain
• Unclear minimum dosing of olive oil and nuts
• The low risk reduction margins may have been accounted for in the baseline differences between the groups including sex, obsesity, use of diuretics, and use of oral hypoglycemics
• Unclear effects of the dietary interventions on cholesterol values
• The incidence of stroke accounts for the reduction in CV events

Funding

• Funded mostly through public sources. Various food makers donated olive oil and nuts, but sponsors were not involved in study design or implementation.