Experimental Outcomes of the Mediterranean Diet: Lessons Learned from the Predimed Randomized Controlled Trial

The Mediterranean Diet (MD) is, culturally and historically, the nutritional pattern shared by people living in the olive-tree growing areas of the Mediterranean basin. It is of great importance for its potential preventive effect against cardiovascular diseases (CVDs). The PREvención con DIeta MEDiterránea (PREDIMED) study, a Spanish multicentre randomised controlled trial (RCT), was designed to assess the long-term effects of the MD, without any energy restriction, on the incidence of CVD in individuals at high cardiovascular (CV) risk. Since its inception, it gave a great contribution to the available literature on the issue. It is well known that, in the field of the health sciences, RCTs provide the best scientific evidence. Thus, the aim of the present review is to analyse the results of the RCTs performed within the frame of the PREDIMED study. Our findings showed that MD has beneficial effects in the primary prevention of CVDs, diabetes and in the management of metabolic syndrome.


Introduction
The Mediterranean diet (MD) is a nutritional model proposed by Ancel Keys, based on the dietary traditions shared around the fifties (1950s) by populations that inhabited the Hellenic peninsula, Italy, and the other countries overlooking the Mediterranean Sea [1]. In descriptive terms, MD is the dietary pattern historically and culturally prevailing among people residing in the olive tree-growing areas of the Mediterranean region before globalization made its effect on lifestyle, diet included [1,2]. Even if the different regions in these areas have their own dietary traditions, they could be considered as variants of the most comprehensive MD [3]. Graphically, it is represented by a pyramid that represents food according to their frequency of intake: rarely to often (weekly or daily), from the basis to the apex, respectively [4].
The MD model is closely related to the history of civilization of the areas surrounding the Mediterranean Sea, and the foods characterizing this dietary pattern have been part of the diet and The main characteristics of the 44 randomized controlled PREDIMED studies and their effects on CVDs and other health outcomes included in our review are shown in Table 1

Neurologic Disorders
Effect of MD on cognition 522 6.5           Effect of MD on telomere lenght 520 5.0 OR (95% CI) for telomere shortening (Δ age adjusted z-score TL ≤ 20th percentile) after 5 years follow-up, adjusted for sex and initial z-score TL:   Effect of MD on telomere lenght 520 5.0 OR (95% CI) for telomere shortening (Δ age adjusted z-score TL ≤ 20th percentile) after 5 years follow-up, adjusted for sex and initial z-score TL:   Effect of MD on telomere lenght 520 5.0 OR (95% CI) for telomere shortening (Δ age adjusted z-score TL ≤ 20th percentile) after 5 years follow-up, adjusted for sex and initial z-score TL:             Changes from baseline at three-monts; differences relative to control diet:

MD + EVOO vs control MD + Nuts vs control
Total cholesterol (mg/dL)   Changes from baseline at three-monts; differences relative to control diet:

MD + EVOO vs control MD + Nuts vs control
Total cholesterol (mg/dL)   Phytosterol intake from natural foods association with a cholesterol-lowering effect of MD   Phytosterol intake from natural foods association with a cholesterol-lowering effect of MD  In the PREDIMED study, a total of 8713 candidates were screened for eligibility, and 7447 of them were enrolled and assigned to one of the three intervention groups (MD + EVOO, MD + Nuts or low-fat diet). Their baseline characteristics according to intervention group are described elsewhere [15]. The exclusion of participants whose randomization procedures were known to have deviated from the protocol did not materially change these results [15]. Participants were followed for a median of 4.8 years (interquartile range: 2. 8-5.8). When compliance with diet intervention was examined, an increase in the 14-item MD questionnaire score was observed for the two MD groups during the follow-up period. Substantial differences between the MD groups and the control group in 12 of the 14 items of the questionnaire were observed. Also, biomarkers' level variations indicated good adherence to the dietary assignments [15]. The main nutrient changes in the MD groups reflected the fat content and composition of the supplementary foods (EVOO or nuts). No relevant diet-related adverse effects were reported. Besides, a little difference in physical activity (assessed with specific questionnaires) among the three groups was observed [15].
As the main objective of the PREDIMED study was to examine the effects of MD on the primary prevention of CVDs, the majority of the RCTs included in our review dealt with CVDs and the related risk factors (Table 1). Estruch et al.'s intention to treat analysis, which included all the 7447 participants, revealed a relative risk reduction of 31% for the MD + EVOO (HR 0.69, 95%CI 0.53, 0.91), and 28% MD + Nuts group (HR 0.72, 95%CI 0.54, 0.95) in the primary composite outcome investigated (including acute myocardial infarction, stroke, or death for CV events), compared to the low-fat control diet group [15]. Moreover, Martínez-González et al., observed that the Hazard Ratio, HR (95% Confidence Interval, CI) for atrial fibrillation in the MD + EVOO group was 0.62 (0.45, 0.85), p < 0.05 [22].
When the effect of MD on diabetes was examined, it was observed that the HR (95% CI) of diabetes incidence was was 0.60 (0.43, 0.85) for the subjects following MD + EVOO compared to controls, and 0.82 (0.61, 1.10) for the MD + Nuts group compared to control diet [36]. After the application of the Fine and Gray model for competing risk analysis, the results remained essentially unchanged [61]. Similarly, a subgroup analysis on the PREDIMED population (n = 418), showed a protective effect of the MD either supplemented with EVOO or nuts against the incidence of DM (HR, 95%CI for both MDs versus control 0.47 (0.26-0.87) [62,63]. Another study showed a significant effect of MD on the incidence of diabetic retinopathy: HR (95% CI) 0.59 (0.37, 0.95) for the MD groups [35].
Further trials evaluated the long-term effect of MD on incidence and reversion of MetS. Although there were no significant differences in incidence or reversion HRs by intervention, reversion occurred in 958 (28.2%) participants when considering only those subjects who had MetS at baseline [38]. Salas-Salvadó et al., examined the one-year effect of the MD on metabolic syndrome (MetS) status, as shown in Table 2. They found that, after 1-year follow-up, the MetS prevalence was reduced by a 6.7%, 13.7% and 2% in the MD + EVOO, MD + Nuts and control groups, respectively (MD + Nuts versus control group, p < 0.05). These differences may be due to the variations in incidence rates among subjects without MetS at baseline and in reversion rates among those who had the syndrome at the beginning of the trial [39].
Álvarez-Pérez et al., [41] found that MD had positive effects on body composition and anthropometric measurements in a subsample of the cohort. Nevertheless, no between-group statistically significant differences were found in anthropometric or body composition variables.
After analysing the influence of a Mediterranean dietary pattern on plasma total antioxidant capacity (TAC), the MD + EVOO group showed higher levels of plasma TAC and a reduction in body weight gain [42].
The effects of the MD on cognitive functions were also examined, as shown in Table 3. In a sub-study conducted on 522 participants in Navarra, it was found that the MD improved cognitive function, assessed with the Mini-Mental State Examination and the clock drawing test [43]. Likewise, another study observed that a long-term intervention with an EVOO-rich MD resulted in a better cognitive function in comparison with controls [44].
Toledo et al.'s study, aimed at investigating the incidence of breast cancer on the PREDIMED population, showed a HR (95% CI) of 0.38 (0.16, 0.87) for the MD + EVOO compared to the control group [53]. Other studies examining the effects of the MD on different conditions, other than CVDs, diabetes obesity and cognitive function, are reported in Table 3.
In order to outline the results obtained by the trials analysed in the present review, we calculated the percentage reduction of the risk of various clinical conditions, as shown in Figure 2. Toledo et al.'s study, aimed at investigating the incidence of breast cancer on the PREDIMED population, showed a HR (95% CI) of 0.38 (0.16, 0.87) for the MD + EVOO compared to the control group [53]. Other studies examining the effects of the MD on different conditions, other than CVDs, diabetes obesity and cognitive function, are reported in Table 3.
In order to outline the results obtained by the trials analysed in the present review, we calculated the percentage reduction of the risk of various clinical conditions, as shown in Figure 2. The % reduction in the risk of cardiovascular disease (a composite of death for cardiovascular cause, non-fatal acute myocardial infarction, and non-fatal stroke) was 31% (95% CI 47-9%) and 28% (95% CI 46-5%) for MD + EVOO and MD + Nuts groups, respectively [15]. Nevertheless, it is appropriate to observe that, although the % risk of CVD reduction vary according to the dietary intervention, it is not possible to infer that one is better than the other, as shown by the overlapping of the correspondent 95% confidence intervals. For the heart failure (HF), the % reduction observed was not significant in the MD + EVOO nor in the MD + Nuts [21], that is to say, none of the two dietary interventions turned out to be better than the control diet in the risk reduction of the outcome. For the atrial fibrillation the % risk reduction was 38% (95% CI 55-15%) for the MD + EVOO group, while not significant for the MD + Nuts group [22]. The risk reduction of peripheral artery disease was 68% (95% CI 81-44%) and 49% (95% CI 68-17%) for the MD + EVOO and MD + Nuts groups, respectively [29], but the difference between the two dietary interventions was not statistically significant due to the partial overlapping of the 95% CIs. For the probability of remaining free of the glucose-lowering medications, a reduction of 22% (95% CI 38-2%) was observed for the MD + EVOO; no significance was observed for the MD + Nuts group [33]. The reduction in the risk of diabetic retinopathy was significant only for the MD + EVOO group (43%, 95% CI: 67-2%) but not for the MD + Nuts group [35]. Interestingly, the long-term effect of MD on diabetic nephropathy was not beneficial, probably due to the higher salt intake than a hyposodic diet (Table 2) [35]. For the incidence of diabetes mellitus, the risk reduction was 40% (95% CI 57%, 15%) and 18% (39%, −10%) for the MD + EVOO and MD + Nuts intervention groups respectively [36], and the difference between the two dietary approaches did not turn out to be statistically significant. For the depression risk, the MD supplemented with either EVOO or Nuts did not lead to a significant reduction, compared to the control diet. However, a risk reduction was observed in the Nuts + MD group among the diabetic subjects only [45]. Finally, the % reduction in the risk of breast cancer incidence was 68% (87-21%) The % reduction in the risk of cardiovascular disease (a composite of death for cardiovascular cause, non-fatal acute myocardial infarction, and non-fatal stroke) was 31% (95% CI 47-9%) and 28% (95% CI 46-5%) for MD + EVOO and MD + Nuts groups, respectively [15]. Nevertheless, it is appropriate to observe that, although the % risk of CVD reduction vary according to the dietary intervention, it is not possible to infer that one is better than the other, as shown by the overlapping of the correspondent 95% confidence intervals. For the heart failure (HF), the % reduction observed was not significant in the MD + EVOO nor in the MD + Nuts [21], that is to say, none of the two dietary interventions turned out to be better than the control diet in the risk reduction of the outcome. For the atrial fibrillation the % risk reduction was 38% (95% CI 55-15%) for the MD + EVOO group, while not significant for the MD + Nuts group [22]. The risk reduction of peripheral artery disease was 68% (95% CI 81-44%) and 49% (95% CI 68-17%) for the MD + EVOO and MD + Nuts groups, respectively [29], but the difference between the two dietary interventions was not statistically significant due to the partial overlapping of the 95% CIs. For the probability of remaining free of the glucose-lowering medications, a reduction of 22% (95% CI 38-2%) was observed for the MD + EVOO; no significance was observed for the MD + Nuts group [33]. The reduction in the risk of diabetic retinopathy was significant only for the MD + EVOO group (43%, 95% CI: 67-2%) but not for the MD + Nuts group [35]. Interestingly, the long-term effect of MD on diabetic nephropathy was not beneficial, probably due to the higher salt intake than a hyposodic diet (Table 2) [35]. For the incidence of diabetes mellitus, the risk reduction was 40% (95% CI 57%, 15%) and 18% (39%, −10%) for the MD + EVOO and MD + Nuts intervention groups respectively [36], and the difference between the two dietary approaches did not turn out to be statistically significant. For the depression risk, the MD supplemented with either EVOO or Nuts did not lead to a significant reduction, compared to the control diet. However, a risk reduction was observed in the Nuts + MD group among the diabetic subjects only [45]. Finally, the % reduction in the risk of breast cancer incidence was 68% (87-21%) for the MD + EVOO group versus the low-fat control diet, while the MD + Nuts did not show to be statistically significant compared to the control group [53].
Overall, the MD + EVOO dietary intervention seemed to have more beneficial effects in terms of % reduction of the risk of different clinical condition. However, in those conditions where both MD + EVOO and MD + Nuts had significative effects compared to the control diet, it is not possible to conclude that the former is better than the latter. Table 4 shows the percentage reduction from baseline of different continuous variables assessed by the different randomized controlled trials conducted in the scope of the PREDIMED study.

Discussion
The RCTs conducted within the frame of the PREDIMED study are the study designs able to best describe the effects of the MD on CVDs and other secondary health outcomes, in terms of sample size, duration of the intervention and follow-up. Nevertheless, in a comprehensive review evaluating the epidemiological and molecular aspects of the MD for non-PREDIMED articles, it was emphasized that only few of them evaluated hard endpoints, and that most of the studies had a sample size smaller than 200 people [1]. It was specified that the most convenient study in terms of number of participants, duration of the intervention and number of publications produced was the PREDIMED study [1]. In the present review, 44 RCTs of PREDIMED study met our inclusion criteria, and the majority of them presented a sample size larger than 200 subjects. The aim of the present review is to summarize the results of RCTs in the PREDIMED study, mainly related to cardiovascular diseases, diabetes, obesity, metabolic syndrome and many other important conditions, and to synthetize the best evidence available.
The results of the PREDIMED study reported in 2013 have been partially retracted due to protocol deviations, mainly regarding the randomization process. Nevertheless, after re-analyzing the collected data with the appropriate corrections (omitting 1588 participants whose study group assignment was known or suspected to have deviated from the protocol), the results obtained were similar [15].
When both the MD groups (MD + EVOO and MD + Nuts) were examined, the MD nutrition model used in the PREDIMED study turned out to potentially reduce the number of hard clinical events in a relatively short time [18]. Firstly, in 2013 it was reported that both intervention groups showed approximately a 30% reduction in the rate of major CV events (myocardial infarction, stroke or death for CV causes), compared to the control group, after a median follow-up of 4.8 years [13].
The epidemiological evidence of the CVD protection provided by the adherence to the MD is strong. A meta-analysis by Liyanage et al., found that the MD was associated with a 37% relative reduction (p < 0.001) in the risk of major CV events [64]. These findings are in agreement with the results of the trials included in the present review, which showed positive effects of the MD on atrial fibrillation [22], and peripheral artery disease [29]. The underlying mechanisms of protection against CVD provided by the MD can be attributed to the abundance of antioxidant and anti-inflammatory molecules in its individual components such as fruits and vegetables, olive oil, nuts, whole grains, fish and red wine, although the specific protective mechanisms of MD on CVDs are not completely understood. One of the hypotheses suggests a possible role of the cell redox state in the modulation of the enzymatic systems related to the antioxidant capacity. Additionally, nutrients have the ability to regulate gene expression and protein synthesis. As reported by nutrigenomic studies, MD can play a role against the expression of several proatherogenic genes involved in vascular inflammation, foam cell formation and thrombosis [18].
As secondary endpoints of the PREDIMED study, diabetes incidence and MetS status were also assessed. The largest trial on the incidence of type 2 diabetes mellitus (T2DM) in the primary prevention PREDIMED study, reported a significant reduction of the incidence in both the intervention groups [36]. Moreover, the results of prospective cohort studies contributing to estimate T2DM risk according to different levels of MD adherence provided additional and consistent evidence [65]. Their results support the protective role of the MD against T2DM, with overall risk reductions ranging from 12% to 83% for subjects closely adhering to the MD compared to those reporting the lowest adherence, after adjusting for several confounders [65]. The authors also observed that higher adherence to the MD had a beneficial role in the prevention and treatment of MetS and its components [65]. In the PREDIMED study, although no differences in the onset of MetS were observed among the three groups, participants in the MD + EVOO and MD + Nuts were more likely to present disease reversion, if compared to the control group [38]. Esposito et al., (2015) specified that two meta-analyses assessed the relationship between adherence to a MD and future incidence of diabetes. According to their report, the analyses are consistent with a significant reduction, ranging from 19% to 23%, of new diabetes diagnosis associated with greater adherence to the MD [66]. In the Framingham Heart Study Offspring Cohort, 1918 participants free of the condition at baseline were followed for seven years, and participants in the highest quintile category of the Mediterranean-style dietary pattern score had a lower incidence of metabolic syndrome than those in the lowest quintile category (p = 0.01) [67]. It is thought that highly important bioactive components of the MD such as unsaturated fatty acids, complex carbohydrates and fibre, vegetable protein, non-sodium minerals, phytosterols and polyphenols interact synergistically to advantageously affect various metabolic pathways at risk of MetS, T2DM and CVD [65].
The role of MD in the protection against cognitive decline, is being supported by growing evidence. Although the majority of the available studies in the issue present a longitudinal or a cross-sectional design, they point out the protective role of MD on cognitive impairment, cognitive function and decline [68].
Among the secondary outcomes of the PREDIMED study, the incidence of breast cancer was assessed. To date, the evidence on the role of Mediterranean diet in the onset of this neoplasm is still limited; nevertheless, the findings of Toledo et al. 's study (2015) are in agreement with the available literature [69,70], and are statistically strengthened by its prospective, randomized and controlled design.
As a result, with the exception of the PREDIMED study, most of the studies on MD appear to be observational studies or short-term trials. Among many issues, the findings of the PREDIMED study include a large number of randomized controlled trials that provide a higher level of scientific evidence than cohort studies and represent the gold standard to clarify the actual effects of this intervention. The PREDIMED trial is a milestone of nutrition intervention that indicated with powerful evidence the benefits of the traditional MD in the primary prevention of CVD in individuals at high risk. As secondary endpoints of the PREDIMED study, it was observed that MD interventions could protect against diabetes in participants without diabetes and figure out a role in preventing or managing MetS. and certain metabolic abnormalities that predicts diabetes and cardiometabolic risk.

Conclusions
In conclusion, the contribution of the PREDIMED study as a commendable dietary intervention study is certain. This trial present as primary endpoint a composite of CV events and, in the frame of the study, sub-group analyses have been performed to assess various secondary outcomes. The scope of this review was to sum up the experimental outcomes of those studies. Randomized controlled trials within the scope of the PREDIMED study demonstrated the risk-reducing effects on major health problems and risk factors as well as the current and known effects of the Mediterranean diet. When the diet is considered as the main determinant of many health outcomes, we testify the Mediterranean diet as a comprehensive diet model that overcomes a single food or single nutrient approach.