The present study is a cross-sectional assessment conducted within the PREDIMED-Plus project (Spain) (www.predimedplus.es). This intervention study aims to evaluate the effect of an intensive intervention with weight loss objectives based on the consumption of a low-calorie MedDiet, promotion of physical activity and behavioral therapy in the primary prevention of CVD and has been described in detail elsewhere . Briefly, the participants included in this project were men (55–75 years) and women (60–75 years) with overweight or obesity (BMI 27–40 kg/m2) who meet at least three criteria of the Metabolic Syndrome (MetS) according to the updated criteria of the International Diabetes Federation and the American Heart Association and National Heart, Lung and Blood Institute  and without prior cardiovascular events.
Recruitment of participants took place between September 2013 and December 2016 including 6874 participants who were randomized. After excluding participants with missing data for the dietary baseline information, for the parameters necessary to the calculation of MetS z-score and those with implausible values for the mean daily energy intake (< 500 and > 3500 kcal/day for women, < 800 and > 4000 kcal/day for men) , 6439 participants were included in the present study (Fig. 1). All participants signed the informed consent, and the project protocol was approved by the Research Ethics Committees from all recruiting centers according to the ethical standards of the Declaration of Helsinki. The trial was registered at the International Standard Randomized Controlled Trial (ISRCTN:http://www.isrctn.com/ISRCTN89898870).
Dietary assessment and pro-vegetarian food patterns
To obtain the final score of the different PVG patterns, the dietary information was evaluated using a semi-quantitative food frequency questionnaire (FFQ) previously validated in Spain [21, 22]. The FFQ was completed at a baseline visit with the help of a trained interviewer. The FFQ includes a list of 143 foods specifying the standard size or ration of consumption over a period of the previous year including 9 possible responses to determine the frequency of consumption ranging from “never or < 1 month” to “ ≥ 6 times a day”.
For the creation of the gPVG pattern, the methodology proposed by Martínez-González  was followed. For healthful and unhealthful PVG versions, the method proposed by Satija et al.  was the reference. Dietary information from 18 food groups (Vegetables, Fruits, Legumes, Whole Grains, Refined Grains, Cooked or Roasted Potatoes, Chips, Nuts, Olive Oil, Tea and Coffee, Fruit Juices, Sugary Drinks, Sweets and Desserts, Meat and Meat Products, Animal Fats, Eggs, Fish and Seafood and Dairy) was used. Table 1 specifies the items included in the 18 food groups and the scoring criteria for each pattern.
In short, to create the different PVG food patterns, consumption in grams of the 18 food groups was adjusted for total energy intake following the residual method . After that, calorie-adjusted consumption in grams was categorized into quintiles giving values of 1–5 according to the consumption quintile of each food group. In the case of the gPVG food pattern seven components, belonging to the plant food groups, scored positively: vegetables, fruits, legumes, grains (whole and refined), potatoes (cooked, roasted and/or fried), nuts and olive oil, and five components (meat and other products, animal fats, eggs, seafood, and dairy), belonging to animal food groups were scored reversely (a value of 5 for lowest consumption). For the hPVG and uPVG, the grain group was separated into whole and refined grains and the potatoes group in fried or chips and cooked or roasted. Four new groups (tea and coffee, natural fruit juices, sweetened drinks and desserts or sweets) were also introduced in both, hPVG and uPVG. To obtain the score of each participant, the points for the 12 components, in the case of the gPVG pattern, and for the 18 groups, in the case of the hPVG and uPVG patterns, were be sum. So, the possible results ranged from 12 points (minimum adherence) to 60 points (maximum adherence) for the gPVG pattern, and from 18 points (minimum adherence) to 90 points (maximum adherence) for the hPVG and uPVG patterns.
MetS z-score and its components
The continuous cardiometabolic risk scale that we used was the MetS z-score proposed by Franks . Prior to the calculation of this scale, all variables were standardized for the total number of participants, except for HDL and waist to hip ratio (WHR) which were standardized by sex using sex-specific cut-off points. The original version of MetS z-score includes fasting insulin in the formula, but we exclude that parameter from the calculation since it was not measured and determined. We also calculated standardized components of MetS z-score (BMI, WHR, SBP/DBP, HDL-c, plasma triglycerides and plasma glucose).
Weight, height, waist and hip circumference were measured by duplicated with light clothing and no shoes using a calibrate scale, a wall-mounted stadiometer, and a non-elastic tape, respectively. Waist circumference was measured midway between the lowest rib and the iliac crest. Hip circumference was measured at the widest part. BMI was calculated as weight (kg) divided by height (meters) squared, and WHR as waist circumference (in cm) divided by hip circumference (in cm). Blood pressure was measured three times with a validated semiautomatic oscillometer after 5 min of rest in-between measurement (Omron HEM-705CP, Hoofddorp, The Netherlands), and the mean of the three measurements was used. After an overnight fast, blood samples were collected at baseline and aliquots of serum and ethylene diamine tetraacetic acid (EDTA) plasma were immediately processed, coded and stored at − 80 °C in a central laboratory until analysis. High Density Lipoprotein (HDL), serum glucose and triglyceride levels were determined by standard enzymatic methods in automatic analyzers in local laboratories.
The MetS z-score for each participant was obtained using the following formula:
(BMI + WHR)/2 + (SBP + DBP)/2 + hyperglycemia (plasma fasting glucose)—HDLc + triglycerides
Other sociodemographic variables, lifestyles and previous history of various diseases, as well as assigned intervention, was also collected at baseline. Information about total physical activity in Metabolic Equivalents (METS) min/day was measured using the validated Regicor Short Physical Activity Questionnaire . Adherence to MedDiet was valued with a 17-item questionnaire, a modified version of a previously validated 14-item questionnaire , for an energy-restricted version.
Descriptive analysis of participants’ characteristics according to quintiles of each PVG food pattern adherence was shown as mean and standard deviation (SD) for quantitative traits, and percentage for categorical variables. We performed the ANOVA test for quantitative variables and the Chi-square test for qualitative variables to compare the characteristics of the sample between adherence quintiles.
Multiple robust linear regression models were performed using an MM-type estimator by adjusting for possible confounders to explore the association between adherence to each PVG food pattern (in quintiles and per 5 points increment in adherence) and MetS z-score, along and with its components separately . Regression coefficients represent the change in each outcome, where 1 unit is equivalent to a 1-SD difference in z scores, or a 1-unit difference in the MetS z-score or its components, per one point of dietary adherence to PVG food patterns, either in the continuous (per each 5 points of adherence) or quintiles form of the different PVG food patterns.
Possible confounder selection was based on a previous review of the literature. It was also adjusted by those variables that when estimating the effect of exposure, the effect changed by ≥ 10% when excluding the variable from the model. Crude model was minimally adjusted for energy intake. Model 1 was additionally adjusted for age (continuous) and sex. Model 2 was additionally adjusted for educational level (illiterate or primary education, secondary education, academic or graduate, and missing information), smoking status (current smoker, former smoker, and never smoker), alcohol intake (grams/day) and total physical activity per day (METS-min/day).
Statistical analyses were carried out with R 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org). For robust linear regression analyses, we also used a robust base package of statistical software R. We used the database version of the PREDIMED-Plus dated March 2019.