Sample population

Participants were recruited from Emory Healthcare outpatient clinics and by advertisements. Assessments were conducted by trained study staff at the Emory University Hospital Clinical Research Center (CRC), formerly the General Clinical Research Center, of the Atlanta Clinical and Translational Science Institute. The study was approved by the Emory University Institutional Review Board and all participants provided written informed consent.

The inclusion criteria were as follows: aged 40–65 years, body mass index (BMI) ≥28 kg/m² and < 35 kg/m², stable body weight in the past 6 months (defined as weight change < 2 kg), currently a non-smoker, currently consuming a stable habitual diet as assessed by interview with the research nutritionist, not consuming anti-oxidants or vitamin-mineral preparations in the past 4 weeks, and 24-h diet recall showing saturated and trans fat intake of > 10% of total daily calories and total cholesterol intake of > 300 mg/day. Participants meeting any of the following criteria were excluded: known history of cardiovascular diseases including coronary artery disease, valvular heart disease, arrhythmias or cardiomyopathies; renal or liver disease defined as creatinine > 2.0 mg/dL and liver function tests > 3-times upper limit of normal; history of diabetes or fasting plasma glucose (FPG) > 126 mg/dL; history of cancer other than skin cancer; systolic blood pressure > 180 mmHg and diastolic blood pressure > 110 mmHg; hematocrit < 30%; and other known acute or chronic illness, including psychiatric disorders, excessive chronic alcohol consumption (> 2 alcoholic beverages/day), statin or other hypolipidemic therapy, and pregnant or lactating females. No imaging was performed for exclusion of fatty liver disease.

Following a screening visit, participants were randomly assigned to one of three groups: (1) Mediterranean diet; (2) habitual high-fat American-type diet supplemented with fish oil, walnuts, and grape juice daily; or 3) habitual high-fat American-type diet (controls). Study measurements were made at a baseline visit and again at 4 and 8 weeks after the baseline visit.

Mediterranean diet intervention

Participants randomized to the Mediterranean diet arm received three meals with beverages and two snacks per day as a prototypical Mediterranean diet prepared by the CRC metabolic kitchen for four consecutive weeks, coupled with verbal and written dietary instruction. During the first four-week period, meals and snacks were picked up by participants at the CRC metabolic kitchen every three to four days and any issues with the meal plan were discussed with the research nutritionist. The seven-day menu from the first week was rotated on a day-to-day basis during the subsequent three weeks to avoid monotony. Verbal and written dietary instruction on the advantages and composition of the Mediterranean diet were given by the CRC research nutritionist during the first four-week period. During the second four-week period, participants received intensive verbal and detailed written dietary instruction and reinforcement to ensure consumption of a prototypical Mediterranean diet using their own home-cooked meals. The CRC research nutritionist formally discussed and reinforced dietary principles with the participant weekly, either in person (baseline, weeks 2, 4, and 6) or via telephone (weeks 1, 3, 5, and 7). Appropriate adjustments were made, as needed, to individual food items to ensure high compliance.

Meals and food plans were designed using ProNutra™ (Viocare, Inc., Princeton, NJ) to provide daily energy for weight maintenance as determined by the Harris-Benedict equation [12]. Protein intake was provided at the Recommended Dietary Allowance level of 0.8 g/kg/day; saturated and trans fats at < 7% of total energy intake; and cholesterol at < 200 mg/day. Using ProNutra™, meals for individual participants were based on the diet scoring for Mediterranean diet adherence [13]. Meals included an abundance of plant food (fruits, vegetables, whole grains, nuts, and legumes); olive oil as the primary source of fat; fish, poultry, and eggs in moderate to low amounts; low consumption of red meats, saturated fats, and sweets; and consumption of either one to two ounce drinks per day of wine (in those who habitually consumed wine) or grape juice (in those who did not habitually consume alcohol).

Diet supplementation intervention

Participants assigned to their habitual high-fat American-type diet supplemented with key components of the Mediterranean diet were given an eight-week supply of specific dietary supplements by the CRC nutritionist and advised to consume them daily, in addition to their usual diet. They were also advised to decrease caloric intake from usual dietary constituents (e.g., not the dietary supplements) if they observed body weight gain of more than 0.5 kg during any given week. The supplements included: (1) two 1-g pills of Omacor® fish oil supplements (1.8 g of EPA/DHA) per day, (2) 1/3 cup shelled walnuts per day, and (3) 16 ounces (about 475 mL) Welch’s® 100% Concord grape juice per day.

Control group

These participants did not alter their diet, take supplements, alter usual activity, and were not given dietary advice.

Dietary intake assessment

Participants were asked to complete 3-day food records at baseline and again at 4 and 8 weeks after the baseline visit. Nutrient composition was determined using ProNutra™ (Viocare, Inc., Princeton, NJ). The 3-day average nutrient intake at each time point was used in the final analysis.

Outcome assessment

Anthropometric measurements (weight and waist circumference) were assessed using standardized procedures by trained CRC staff. Blood lipids (total cholesterol, triglycerides, high-density lipoprotein [HDL] cholesterol, and low-density lipoprotein [LDL] cholesterol) were measured using a CX7 chemistry analyzer (Beckman Diagnostics, Fullerton, CA) by technicians at the Emory Healthcare Medical Laboratory. Inflammatory biomarkers and adipokines included IL-6, IL-8, C-reactive protein (CRP), and adiponectin. IL-6 and IL-8 were measured with a Fluorokine MAP MultiAnalyte Profiling Human Base Kit (R&D Systems, Inc., Minneapolis, MN) on a Luminex – 200 platform. IL-6 values that were below the limit of detection were assigned a value of 0.01 pg/mL. CRP was measured using the Dade-Behring Nephelometry System (BNII). One participant had a CRP level less than the limit of detection (LOD) and was assigned a value of LOD/sqrt(2). Adiponectin was measured with the Quantikine Human Adiponectin Immunoassay solid-phase ELISA (R&D Systems, Inc., Minneapolis, MN). Markers of hyperglycemia, including FPG and insulin were measured using a CX7 chemistry analyzer (Beckman Diagnostics, Fullerton, CA) by technicians at the Emory Healthcare Medical Laboratory.

Brachial artery reactivity was used to test flow-mediated vasodilation (FMD) as a measure of endothelial function and more specifically of nitric oxide bioavailability. This outcome was chosen because endothelial dysfunction integrates risk factor-mediated injury to the endothelial cells. It is an early marker of risk for development of atherosclerosis and its adverse outcomes and thus provides a sensitive, reproducible and non-invasive tool for investigation of the effects of short-term dietary intervention (e.g., an excellent experimental tool) [14, 15]. All measurements were performed in a temperature-controlled laboratory after an overnight fast. The brachial artery of the non-dominant arm was imaged using a high-resolution 10 MHz linear array ACUSON™ ultrasound transducer (Siemens Medical Solutions USA, Inc., Malvern, PA) at baseline and continually for 120 s after producing 5-min ischemia of the hand. Arterial diameter was measured as the distance from the leading edge of the intima-lumen interface of the near wall to the leading edge of the lumen-intima interface of the far wall by an investigator blinded to the treatment. FMD was calculated as the percent increase in brachial artery vasodilator response: (post-hyperemia diameter-baseline diameter)/baseline diameter × 100.

Circulating pro-angiogenic cell activity (CFU-As) were measured using a colony forming assay from circulating mononuclear cells as described previously [16, 17]. Briefly, mononuclear cells were isolated by density-gradient centrifugation from a 20 ml sample of venous blood using CPT tubes (Becton, Dickinson and Company, Franklin Lakes, NJ), and washed two times with PBS. The cells were suspended in growth medium (DMEM supplemented with 20% fetal bovine serum and 6.5% endothelial cell growth supplement), and plated on human fibronectin-coated cell culture dishes. To eliminate mature circulating endothelial cells, cells adherent after 24 h were discarded and nonadherent cells were re-plated onto new fibronectin-coated plates at 1 million cells/well. Growth medium was changed every 2 days. After 7 days, CFU-As were counted manually and recorded by an observer who was blinded to the clinical data. Colonies were identified as central clusters of rounded cells with multiple flat cells emanating from the central clusters.

Biomarkers of oxidative stress included plasma cysteine (reduced form), cystine (oxidized form), and glutathione (reduced form), and were measured using high-performance liquid chromatography with fluorescence detection as described previously [18].

Statistical analysis

Values presented are mean with standard deviation given in parentheses. Normality was tested using the Shapiro-Wilk test statistic and visually assessed using Q-Q plots. Differences in demographic, clinical, and dietary factors between arms at baseline were assessed using Fisher’s exact test for categorical variables and analysis of variance (ANOVA) for continuous variables. Within-subject differences in the health outcomes were analyzed using Student’s paired t-test for normally distributed variables and Wilcoxon sign ranks test for non-normally distributed variables [19, 20]. Mixed-effects models were used to analyze between-subject differences in the health outcomes [21]. Models included a group-by-time interaction term. Baseline health outcome was adjusted in all models. Statistical significance was considered for P < 0.05. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC).

Change in dietary intake

Effects of Mediterranean diet

Among participants who were assigned to the Mediterranean diet arm, there was a significant decrease from baseline to 8 weeks in body weight (− 2.2 [2.6] kg, p = 0.03), total cholesterol (− 24.9 [19.8] mg/dl, p = 0.003), LDL cholesterol (− 24.9 [17.3] mg/dl, p = 0.003), and CFU-As (− 40.3 [36.4] cfu, p = 0.01). There was also a non-significant increase in FMD from baseline to 4 weeks (2.3 [4.0] percent, p = 0.08). The average percent increase in FMD relative to baseline among participants assigned to the Mediterranean diet arm was 66.1% at 4 weeks and 44.3% at 8 weeks.

When compared to changes in controls, participants in the Mediterranean diet arm had significantly greater weight loss at both 4 and 8 weeks and greater decreases in total cholesterol and LDL cholesterol at 4 weeks (Table 1; Fig. 1). Moreover, compared to controls, participants in the Mediterranean diet arm had significantly greater decreases in CFU-As at 8 weeks, and in plasma cystine (indicative of decreased oxidative stress) at both 4 and 8 weeks. However, compared to changes in controls, participants in the Mediterranean diet arm had no significant changes in any of the inflammatory biomarkers or adipokines (IL-6, IL-8, CRP, and adiponectin), markers of hyperglycemia (FPG and insulin), or FMD. In a subgroup analysis of participants with FMD < 6% at baseline (n = 8 Mediterranean diet arm and n = 5 controls), there was a non-significant increase in FMD (p = 0.07) among those in the diet group compared to controls at 4 weeks.

Effects of dietary supplementation with fish oil, walnuts, and grape juice

Among participants who received supplements, there was a significant decrease from baseline to 8 weeks in triglycerides (− 17.4 [16.8] mg/dl, p = 0.01), but in none of the other measures. When compared to changes in controls, the supplements resulted in significantly greater decreases in adiponectin levels at 4 weeks, but no significant changes in weight, waist circumference, lipid levels, inflammatory or oxidative stress biomarkers, markers of hyperglycemia, FMD, or CFU-As (Table 1; Fig. 1).