Mediterranean-Style Diet for the Primary and Secondary Prevention of Cardiovascular Disease: A Cochrane Review

Background: Diet plays a major role in cardiovascular disease (CVD) risk. Objectives: To determine the effectiveness of a Mediterranean-style diet for the primary and secondary prevention of CVD. Methods: We searched for randomised controlled trials (RCTs) of Mediterranean-style diets in healthy adults and those at increased risk of CVD (primary prevention) and with established CVD (secondary prevention). Results: Thirty RCTs were included, 22 in primary prevention and eight in secondary prevention. Clinical endpoints were reported in two trials where there was moderate quality evidence for a reduction in strokes for primary prevention, and low quality evidence for a reduction in total and CVD mortality in secondary prevention. We found moderate quality evidence of improvement in CVD risk factors for primary prevention and low quality evidence of little or no effect in secondary prevention. Conclusions: There is still some uncertainty regarding the effects of a Mediterranean-style diet in CVD prevention.


Introduction
Cardiovascular disease (CVD) is currently the leading cause of mortality worldwide, causing one-third of deaths globally [1]. Preventative strategies focus on modifiable risk factors such as high blood pressure, high cholesterol, smoking, obesity and poor diet, which account for a large proportion of the overall CVD burden [1], with poor diet estimated to be responsible for the largest contribution in Europe [2].
The Seven Countries study from the 1960s demonstrated the potential beneficial effects of a Mediterranean dietary pattern [3]. Systematic reviews of prospective observational studies have confirmed that greater adherence to a Mediterranean diet is associated with a significant improvement in health status and a significant reduction in overall mortality, as well as in morbidity and mortality from CVD and other major chronic diseases [4][5][6][7][8][9]. Benefits have also been found for individual CVD risk factors [10,11]. There is less evidence however from well-conducted and adequately powered randomised controlled trials (RCTs).
Studies also differ in their definition of a Mediterranean diet. The original description reflects the common dietary pattern of communities in countries of the Mediterranean region in the early 1960s [3], which was an expression of common cultural and historical roots, and a shared set of lifestyle and eating habits transferred data into the Review Manager software [31]. We double-checked that data were entered correctly by comparing the data presented in the systematic review with the data extraction form.

Assessment of risk of bias in included studies
Two review authors (of KR, LE, DW, AV, AD, LH) independently assessed risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions [32]. We resolved any disagreements by discussion or by involving another author (KR). We assessed the risk of bias according to the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting and other bias. We graded each potential source of bias as high, low or unclear.

Measures of treatment effect
We processed data in accordance with the Cochrane Handbook for Systematic Reviews of Interventions [32]. We expressed dichotomous outcomes as risk ratios (RR) with 95% confidence intervals (CI). Where available we used adjusted estimates of treatment effect as hazard ratios, and used the inverse variance method to pool these statistically. For continuous variables, we compared net changes (i.e. intervention group minus control group differences) and calculated mean differences (MD) and 95% CIs for each study. For trials with multiple arms we divided the control group N by the number of arms to avoid double-counting in metaanalyses. We analysed outcomes at the longest period of follow-up where multiple measurements had been taken unless there was significant (> 30%) attrition. We narratively described skewed data reported as medians and interquartile ranges.

Data synthesis
We summarised and analysed all eligible studies in Review Manager 5. We undertook meta-analyses only where this was meaningful, i.e. if the treatments, participants and the underlying clinical question were similar enough for pooling to make sense. We used a random-effects model as we could not assume that all studies in the meta-analysis are estimating the same intervention effect, but rather are estimating intervention effects that follow a distribution across studies. We used the I² statistic to measure heterogeneity among the trials in each analysis. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of a body of evidence as it relates to the studies that contribute data to the meta-analyses for the prespecified outcomes. In the main analysis we did not combine primary and secondary prevention studies and different comparator groups as this would have made interpretation of the results difficult due to heterogeneity; instead we conducted four main analyses: Mediterranean dietary intervention versus no intervention or minimal intervention for primary prevention; Mediterranean dietary intervention versus another dietary intervention for primary prevention; Mediterranean dietary intervention versus usual care for secondary prevention; Mediterranean dietary intervention versus another dietary intervention for secondary prevention.

Study selection and characteristics
Searching for this updated Cochrane review yielded 9483 records after duplicates were removed. Following screening of titles and abstracts 9296 were excluded and 187 full text papers were assessed for eligibility. Twenty one completed and seven ongoing trials were included from the new search. Six studies were included from the previous review and rescreening the original database identified three further trials, so a total of 30 completed studies and seven ongoing trials were included in this review (Figure 1).
The characteristics of the 30 completed included studies are shown in Table 1. Results are described for each of the four main analyses.

Mediterranean dietary intervention versus usual care for secondary prevention
Two trials were included with 706 participants randomised. Both trials recruited patients with CVD, one in men and women with CHD [48], and the other in men and women who had experienced a myocardial infarction within six months [58]. Participants were recruited from Germany [48], and France [58]. The duration of the intervention and follow-up periods varied from 12 months to 24 and 46 months [48,58].

Mediterranean dietary intervention versus another dietary intervention for secondary prevention
Six trials were included with 1731 participants randomised. An expression of concern has been published about the reliability of two of the studies in this comparison group with the majority of participants [53,54,63,64], and we have conducted sensitivity analyses excluding these studies from all analyses. All trials recruited patients with CVD. Three trials recruited men and women with CHD [38,47,62], one after a first myocardial infarction [59] and one with acute myocardial infarction or unstable angina [53]. One trial recruited patients with established CHD or those at high risk of CHD, although the majority of participants had established disease [54]. Participants were recruited from Australia [38,47], the US [59], Brazil [62], and India [53,54]. The duration of the intervention and follow-up periods varied: three months [38,62], six months [47], and two years [53,54,59]. The dietary interventions in the comparison group varied, including low-fat [38,47,59], and national recommendations/disease-specific guidance [53,54,62].

Risk of bias in included studies
The risk of bias of included studies is shown in Figure 2. The blinding of participants and personnel for behavioural interventions is difficult, if not impossible, in most cases and so we have not judged this as a high risk of bias. We rated this domain as unclear for all trials in all four comparison groups.

Mediterranean dietary intervention versus no intervention or minimal intervention for primary prevention
The methods of random sequence generation were unclear in six of the nine included studies [35,36,37,41,46,61], and we judged the methods used to be at low risk of bias in the remaining three [39,43,44]. The methods of allocation concealment were unclear in seven studies and at low risk of bias in the remaining two [43,61]. Blinding of outcome assessment was unclear in eight trials, in the remaining trial we judged this to be at low risk of bias [43]. We judged three trials to be at low risk of attrition bias [43,44,61], one study was at high risk of bias as there was differential loss to follow-up [41], and the remaining trials were judged as unclear. The risk of bias associated with selective reporting was judged as low risk in five trials [39,41,43,44,61], and unclear in the remaining four. There was insufficient information to judge the risk of other sources of bias and we categorised all nine studies as unclear.

Mediterranean dietary intervention versus another dietary intervention for primary prevention
The methods of random sequence generation were unclear in six of the 13 included studies [33,45,40,52,55,60], and judged at low risk of bias in the remaining seven [34,42,49,50,51,56,57]. The methods of allocation concealment were unclear in 10 studies and at low risk of bias in the remaining three [42,56,57]. Blinding of outcome assessment was unclear in 10 of the 13 trials and judged at low risk of bias in the remaining three [34,49,56]. We judged six of the 13 trials to be at low risk of attrition bias [33,34,42,49,51,56], one study to be at high risk of bias for attrition due to differential loss to follow-up [60], and unclear in the remaining six trials. The risk of bias associated with selective reporting was unclear in four studies and judged to be of low risk of bias for the remaining nine [33,34,42,49,50,51,55,56,60]. There was insufficient information to judge the risk of other sources of bias and we categorised all 13 studies as unclear.

Mediterranean dietary intervention versus usual care for secondary prevention
The methods of random sequence generation were unclear in one of the two included studies [58], and in the other we judged the methods used to be at low risk of bias [48]. The methods of allocation concealment were unclear in one study [48], and at low risk of bias in the other [58]. Blinding of outcome assessment was unclear in one trial at low risk of bias for the other [48,58]. We judged both trials to be at low risk of attrition bias and selective reporting. There was insufficient information to judge the risk of other sources of bias and we categorised both studies as unclear.

Mediterranean dietary intervention versus another dietary intervention for secondary prevention
The methods of random sequence generation were unclear in five of the six included studies [38,53,54,59,62], and in the one study where this was clear, we judged the methods used to be at low risk of bias [47]. The methods of allocation concealment were unclear in four studies and at low risk of bias in the remaining two [59,62]. Blinding of outcome assessment was unclear in five studies and judged to be at low risk of bias in one [54]. The risk of attrition bias was unclear in all six trials. The risk of bias associated with selective reporting was unclear in three studies and at low risk of bias in three [54,59,62]. An expression of concern has been published about the reliability of two of the studies in this comparison group [53,54,63,64]. We have conducted sensitivity analyses excluding these studies from all analyses. We regarded these two studies as at high risk of other bias. We judged the remaining four studies as at unclear risk of other sources of bias as there was insufficient information to make a judgement.

CVD clinical events
We assessed the overall quality of evidence using the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) presented in Table 2 for each of the four main comparisons.

Mediterranean dietary intervention versus no intervention or minimal intervention for primary prevention
None of the nine included studies reported on clinical events.

Mediterranean dietary intervention versus another dietary intervention for primary prevention
The PREDIMED trial was the only trial reporting clinical events for this comparison [51]. This trial was retracted and re-analysed following concerns regarding randomisation at two of 11 sites and the inclusion of non-randomised second household members. Low-quality evidence shows little or no effect of the intervention (advice to follow a Mediterranean diet plus supplemental extra-virgin olive oil or tree nuts) compared to a low-fat diet on CVD mortality (hazard ratio [HR] 0.81, 95% confidence interval [CI] 0.50 to 1.32) or total mortality (HR 1.0, 95% CI 0.81 to 1.24) over 4.8 years. There was, however, a reduction in the number of strokes with the intervention (HR 0.60, 95% CI 0.45 to 0.80), moderate-quality evidence).

Mediterranean dietary intervention versus usual care for secondary prevention
The Lyon Diet Heart Study examined the effect of advice to follow a Mediterranean diet and supplemental canola margarine compared to usual care in 605 CHD patients over 46 months and there was low-quality evidence of a reduction in adjusted estimates for CVD mortality (HR 0.35, 95% CI 0.15 to 0.82) and total mortality (HR 0.44, 95% CI 0.21 to 0.92) with the intervention [58].

Mediterranean dietary intervention versus another dietary intervention for secondary prevention
Three studies reported clinical endpoints for this comparison group [53,54,59], and two of these were excluded in sensitivity analyses from all main analyses due to published concerns regarding the reliability of the data [53,54,63,64]. Only one small trial (101 participants) provided unadjusted estimates for composite clinical endpoints for comparison four (very low-quality evidence of uncertain effect) [59].

Cardiovascular risk factors
We assessed the overall quality of evidence using the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) presented in Table 3 for each of the four main comparisons. Figures 3-8 show stacked forest plots for the four main comparisons for each of the following outcomes: Total cholesterol (Figure 3), LDL cholesterol (Figure 4), HDL Cholesterol (Figure 5), Triglycerides (Figure 6), Systolic Blood Pressure (Figure 7), Diastolic Blood Pressure (Figure 8).

Mediterranean dietary intervention versus no intervention or minimal intervention for primary prevention
For CVD risk factors for comparison one, there was low-quality evidence for a possible small reduction in total cholesterol (-0.16 mmol/L, 95% CI -0.32 to 0.00) and moderate-quality evidence for a reduction in systolic (-2.99 mmHg (95% CI -3.45 to -2.53) and diastolic blood pressure (-2.0 mmHg, 95% CI -2.29 to -1.71), with low or very low-quality evidence of little or no effect on LDL or HDL cholesterol or triglycerides.

Mediterranean dietary intervention versus another dietary intervention for primary prevention
For comparison 2 there was moderate-quality evidence of a possible small reduction in LDL cholesterol (-0.15 mmol/L, 95% CI -0.27 to -0.02) and triglycerides (-0.09 mmol/L, 95% CI -0.16 to -0.01) with moderate or low-quality evidence of little or no effect on total or HDL cholesterol or blood pressure.   •○○○ VERY LOW Little or no effect *sensitivity analysis without Singh studies.

Mediterranean dietary intervention versus usual care for secondary prevention
For comparison three there was low-quality evidence of little or no effect of a Mediterranean-style diet on lipid levels and very low-quality evidence for blood pressure.

Mediterranean dietary intervention versus another dietary intervention for secondary prevention
For comparison four where only two trials contributed to the analyses there was low or very low-quality evidence of little or no effect of the intervention on lipid levels or blood pressure.

Summary of main results
In this substantive review update, 30 RCTs (49 papers) and seven ongoing trials met our inclusion criteria. Four pre-specified comparison groups were used to analyse the data to address both heterogeneity between participants and comparison groups and aid interpretation of findings. Clinical endpoints were measured in only one large primary prevention trial (PREDIMED) [51], where following retraction and reanalysis of the data due to methodological concerns, there was low quality evidence of little or no effect of the intervention on total or CVD mortality or myocardial infarction but moderate quality evi-dence of a reduction in the number of strokes. For secondary prevention, in the Lyon Diet Heart Study [58], there was low quality evidence of reductions in total and CVD mortality and myocardial infarction. Two secondary prevention trials reporting clinical events were excluded from the main analyses due to published concerns regarding the reliability of the data [53,54,63,64]. A further small trial reported unadjusted estimates for total cardiac endpoints, with very low-quality evidence showing considerable uncertainly of the effect size. Cardiovascular risk factors including lipid levels and blood pressure were reported in all four comparison groups, with most of the studies contributing to primary prevention. Compared to no intervention or a minimal intervention, there was low-quality evidence for a possible small reduction in total cholesterol and moderate-quality evidence for a reduction in both systolic and diastolic blood pressure with a Mediterranean-style diet, with low or very low-quality evidence of little or no effect of the intervention on LDL or HDL cholesterol or triglycerides. Compared with other diets in primary prevention, there was moderate-quality evidence of a possible small reduction in LDL cholesterol and triglycerides with moderate or low-quality evidence of little or no effect of the intervention on total or HDL cholesterol or blood pressure. In secondary prevention few trials contributed to the analyses with low or very low-quality evidence of little or no effect of a Mediterranean-style diet on lipid levels or blood pressure.

Study limitations and strengths
There were a large number of included trials (30 trials, 12,461 participants randomised), but few reported on clinical endpoints, our primary outcome, and the majority of trials reported on CVD risk factors for primary prevention. Due to the breadth of the review question, heterogeneity in terms of participants, interventions and comparators was high and we have attempted to reduce this by conducting the main analyses in four comparison groups for primary and secondary prevention and different comparators. Two studies were excluded from all main analyses in sensitivity analyses due to published concerns regarding the reliability of the data [53,54,63,64]. Only one trial reported clinical endpoints for primary prevention and this study experienced methodological issues regarding randomisation with the report subsequently being retracted and re-analysed [51]. The findings in secondary prevention are based on one older trial reporting very large effect estimates using a modified Zelen design [58]. Overall most trials were at unclear risk of bias for most domains and so results should be treated with caution.
We conducted a comprehensive search across major databases for interventions involving the Mediterranean diet. Two review authors independently selected and assessed trials for inclusion using prespecified criteria, extracted data and assessed the quality of trials to minimise potential biases in the review processes. Our decision to restrict this review to interventions that only focused on the effectiveness of a Mediterranean-style diet per se avoided the potential confounding effects of other behavioural interventions on our outcomes, for example, those involving increased exercise in the context of trials examining multifactorial interventions. Our decision to exclude trials in people with diabetes who are at increased risk of CVD also missed relevant studies, but people with diabetes represent a distinct group and interventions for the management of diabetes are covered by another review group, the Cochrane Metabolic and Endocrine Disorders Group.
Definitions of the Mediterranean diet differed across studies. We used a classification system in our definition in an attempt to address the heterogeneity. The components required to meet our definition of a Mediterranean-style diet were based on previous definitions [13][14][15][16][17], and required at least the following two core components: high monounsaturated/saturated fat ratio (use of olive oil as main cooking ingredient and/or consumption of other traditional foods high in monounsaturated fats such as tree nuts) and high intake of plant-based foods, including fruits, vegetables and legumes. The rationale for this definition is based on recent work [5,30], which emphasises that the protective effects of the diet appear to be most attributable to olive oil, fruits, vegetables and legumes.

Comparison with similar studies
Several recent systematic reviews and overviews of reviews have reported on the effects of the Mediterranean diet on cardiovascular health. A recent narrative overview of both prospective observational studies and RCTs concluded that the Mediterranean diet has some beneficial effects for CVD prevention but the effects were inconsistent between studies with few studies reported in meta-analyses and the overview authors called for more high-quality trials to address the inconsistencies [65]. This is in line with the findings of the current review reporting on RCT evidence. An umbrella review of systematic reviews reported on 13 meta-analyses of observational studies and 16 meta-analyses of RCTs investigating the association between the adherence to the Mediterranean diet and a number of different health outcomes [4]. The authors found robust evidence for a greater adherence to the Mediterranean diet and a reduced risk of overall mortality, cardiovascular diseases, coronary heart disease, myocardial infarction and diabetes with no evidence for LDL cholesterol levels.
A recent systematic review included both primary and secondary prevention trials and pooled clinical endpoints for these showing beneficial effects for major vascular events (risk ratio (RR) 0.69, 95% CI 0.55 to 0.86) and stroke (RR 0.66, 95% CI 0.48 to 0.92) [66]. A systematic review comparing the effects of a Mediterranean diet with low-fat diets on CVD risk factors in those at high risk or with established disease found favourable but modest effects of the Mediterranean diet on a wide range of cardiovascular risk factors and inflammatory markers, such as body weight, systolic and diastolic blood pressure, fasting plasma glucose, total cholesterol and high-sensitivity C-reactive protein [67]. Other systematic reviews have pooled together the evidence from both observational studies and RCTs on the effects of the Mediterranean dietary pattern on metabolic syndrome and individual cardiovascular risk factors, supporting favourable effects of the Mediterranean diet on cardio-metabolic risk factors [10,11]. The results of the current review including only RCTs show inconsistencies between studies but where meta-analyses were possible there were small beneficial effects on some CVD risk factors for primary prevention.

Conclusions
Despite the large number of trials included in the review there is still uncertainty regarding the effects of a Mediterranean-style diet on clinical endpoints and cardiovascular disease (CVD) risk factors for both primary and secondary prevention from current clinical trial evidence. Two trials reporting clinical endpoints for secondary prevention were excluded because of concerns regarding the reliability of the data, so the available evidence is restricted to one large trial and a small trial reporting unadjusted estimates of effect. Evidence for primary prevention on clinical endpoints is limited to one large trial with methodological issues (although these have now been addressed in a recent re-analysis). Further adequately powered primary prevention trials are needed to confirm findings on clinical endpoints to date. Many trials reported on CVD risk factors, particularly in primary prevention, but heterogeneity precluded metaanalyses for some outcomes. Several ongoing trials have been identified which will add to the evidence base.