A provegetarian (PVG) food pattern is reportedly cardioprotective in older adults from non-Mediterranean populations, but scarce evidence is available for Mediterranean populations. We assessed the association between 3 different PVG food patterns and the risk of cardiovascular disease (CVD) in the SUN (Seguimiento Universidad de Navarra) cohort.
MethodsThe SUN Project is a Mediterranean cohort study of relatively young Spanish university graduates. A PVG food pattern was calculated by assigning positive scores to plant-based foods and reverse scores to animal-based foods. Two additional patterns, healthful (hPVG) and unhealthful (uPVG), were derived based on the quality of plant-based foods.
ResultsThe final analysis included 18 560 participants, with a mean±standard deviation baseline age of 38±12 years. During a median [interquartile range] follow-up of 16 [10-20] years, 227 cases of CVD were identified. Participants within the upper quartile of the PVG pattern had a 37% lower CVD risk (HRQ4vsQ1, 0.63; 95%CI, 0.42-0.94) compared with the lowest quartile. This inverse association was similar for the hPVG (HRQ4vsQ1, 0.60; 95%CI, 0.40-0.90), but not for the uPVG food pattern, which instead exhibited a 76% higher CVD risk between extreme quartiles (HRQ4vsQ1, 1.76; 95%CI, 1.17-2.64).
ConclusionsBoth a general PVG and hPVG food pattern were associated with a reduced CVD risk in a relatively young Mediterranean population. An uPVG food pattern was associated with an increased risk.
Keywords
In recent years, an increasing number of individuals have adopted plant-based diets, motivated by concerns regarding animal welfare, environmental impact, and health.1,2 Plant-based diets that do not entirely exclude animal-derived foods are more easily sustained over the long term compared with strict vegetarian or vegan diets. Several studies have reported that plant-based diets, which emphasize the intake of plant-derived foods while not necessarily excluding animal-derived foods, are associated with a reduced cardiovascular disease (CVD) risk.3–9 However, not all plant-based diets are created equal, and their impacts on health can vary significantly.4–6,9–16
There is a scarcity of studies in Mediterranean populations.11,17,18 In a cross-sectional analysis in the European Research Council-funded PREDIMED-Plus randomized intervention trial19 conducted in older adults in a Mediterranean population at high cardiometabolic risk, participants with greater adherence to a healthful provegetarian (hPVG) food pattern exhibited an estimated lower cardiometabolic risk than those with lower adherence. Conversely, individuals with higher adherence to an unhealthful provegetarian (uPVG) food pattern displayed an increased estimated cardiometabolic risk relative to those with lower adherence.17 However, to our knowledge, no large prospective cohort studies have been conducted in younger adults.
The aim of this study was to examine the association between 3 provegetarian food patterns (general, healthful, and unhealthful) and CVD risk in a Mediterranean population of relatively young adults.
METHODSStudy design and populationThe Seguimiento Universidad de Navarra (SUN) Project is an ongoing, prospective, and multipurpose cohort study involving relatively young Spanish university graduates. The design, objectives, and methodology have been widely described elsewhere.20 Recruitment began in December 1999. Participants complete an initial questionnaire, followed by biennial follow-up questionnaires to update information on their lifestyle behaviors and health conditions.
From December 1999 to April 2024, a total of 23 321 participants were recruited (figure S1). We excluded 186 participants with insufficient follow-up, 353 participants with prevalent CVD at baseline, 2276 participants with implausible total energy intake (below or above the 5th and 95th percentiles, respectively), 153 participants who did not provide answers to 70 or more items on the Food Frequency Questionnaire (FFQ), and 1793 participants with no follow-up information (retention rate: 91%). The final sample consisted of 18 560 participants.
Volunteers gave implied consent by completing the baseline questionnaire; the study followed the Declaration of Helsinki and was approved by the Ethics Review Board of the University of Navarre (number 2001_30).
Diet was assessed at baseline and after 10 years of follow-up with a self-administered 136-item semiquantitative FFQ. This FFQ has been previously validated in Spain.21,22
Dietary assessmentThe overall PVG was originally described by Martínez-González et al. in 2014,24 who reported a monotonic inverse and strong prospective association between overall PVG and all-cause mortality in participants of the PREDIMED randomized trial.23 These results were followed by those of several other cohorts. Based on the original PREDIMED finding, in 2016, Satija et al.16 also defined the hPVG and uPVG patterns in the Harvard cohorts. To compute the hPVG and uPVG, we followed the methods used by Satija et al.16 and previously applied in the SUN cohort.25–29
We defined 19 food groups: 8 were healthy plant-origin food groups, 5 were unhealthy plant-origin food groups, and 6 were animal-origin foods. The residual method was used separately for men and women to adjust the intake of each food group (g/d) for total energy intake.30 The energy-adjusted estimates (residuals) were ranked according to their sex-specific quintiles, and each quintile received a score ranging from 1 to 5. To obtain the PVG, plant-origin food groups were assigned positive scores, whereas animal-origin food groups received negative scores. For the hPVG food pattern, only healthy plant-origin foods were given positive scores, while unhealthy plant-origin foods and animal-origin foods were assigned negative scores. In contrast, for the uPVG food pattern, unhealthy plant-origin foods were rated positively, while both healthy plant-origin foods and animal-origin foods received negative scores. The quintile values of the food groups were summed. Their final score could range from 19 (lowest adherence) to 95 (highest adherence). Table S1 lists the food items included in each PVG food pattern with their respective scoring criteria.
Participants were categorized into quartiles based on their adherence to each PVG pattern. The intake of both margarine and alcoholic beverages was excluded from the score in accordance with the methods described by Martínez-González et al.23 and Satija et al.16
Outcome assessmentThe endpoint of this study was the incidence of CVD, which included nonfatal acute coronary syndrome (ST-segment and non–ST-segment myocardial infarction), nonfatal stroke, and cardiovascular death. When CVD was self-reported in any of the biennial follow-up questionnaires, incident CVD diagnoses were confirmed by reviewing participants’ medical records. An expert committee of cardiologists, blinded to participant diet and lifestyle exposure, worked as adjudicators of confirmed cases.
The fourth universal definition of myocardial infarction was applied to nonfatal coronary syndromes.31 Nonfatal stroke was defined as a focal neurological deficit of sudden onset that lasted more than 24hours and had a vascular mechanism. Deaths were reported by next-of-kin, work colleagues, or postal authorities. Cardiovascular deaths were confirmed according to the 10th edition of the International Classification of Diseases via a review of medical records and reports with the permission of the participants’ next-of-kin. The Spanish National Death Index was checked annually to determine the cause of death of cohort members who died during follow-up. Information on the vital status and cause of death of the deceased was provided by the Spanish National Institute of Statistics through a specific agreement.
Assessment of other variablesWe obtained information about sociodemographic characteristics, anthropometric variables, dietary variables, lifestyle factors, history of chronic diseases, use of cardiovascular drugs, and family history of CVD. Self-reported weight and BMI were previously validated in a subsample of this cohort.32 A 9-point adherence score to the Mediterranean diet (MedDiet) was computed following the method of Trichopoulou et al.33
Statistical analysisCox regression models were fit to assess the association between PVG food patterns and CVD risk. Hazard ratios (HR) were calculated with their 95% confidence intervals (95%CI), including age as the underlying time variable and always considering the lowest quartile of adherence to each dietary pattern as the reference category. Entry time was defined as the date of completion of the first questionnaire, while exit time was defined as the date of a cardiovascular event, the date participants completed their last follow-up questionnaire, or the date of death from a cause unrelated to CVD for those who did not experience a CVD event.
The Cox regression models were adjusted for potential confounders. Models were stratified by age (10-year period) and the period of completion of the baseline questionnaire (6-year period). In addition, tests of linear trend among successive quartiles of adherence were conducted.
We conducted repeated measurements to assess the impact of dietary intake variation on participants who completed the FFQ after 10 years of follow-up. Cumulative averages for dietary information were used.
To evaluate the degree of overlap between each PVG food pattern and the MedDiet, we computed Pearson's correlation coefficients between each PVG food pattern and the MedDiet 9-point score.33 We also calculated the proportion of participants who were assigned to the same quartile for both food patterns.
Although we accounted for a broad range of potential confounding variables, we cannot completely exclude the possibility of residual confounding. To address this issue, we computed the E-value proposed by VanderWeele et al.34
We studied possible effect modification of each PVG food pattern by sex, age, BMI, physical activity, and smoking status. In this subgroup analysis, statistical interaction was assessed using likelihood ratio tests.
To assess the robustness of our results, sensitivity analyses were performed by evaluating the models in different scenarios.
Analyses were performed with STATA/SE V15.0 (StataCorp, College Station, United States). A 2-sided P <.05 was considered statistically significant.
RESULTSData from 18 560 participants were analyzed (62% women). The mean±standard deviation age at baseline was 38±12 years. During a median [interquartile range] follow-up of 16 [10-20] years and observation of 234 867 person-years at risk, we registered 227 confirmed incident cases of CVD (acute myocardial infarction (fatal and nonfatal): 96 events; stroke (fatal and nonfatal): 76 events; and CVD deaths (excluding acute myocardial infarction and stroke): 55 events).
The baseline characteristics of participants across quartiles of each PVG pattern are presented in table 1. Table S2 presents baseline dietary information from the SUN participants.
Baseline characteristics of participants according to quartiles of the provegetarian, healthful provegetarian, and unhealthful provegetarian food patterns in the SUN Project
| Variables | PVG | hPVG | uPVG | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Q1 | Q2-Q3 | Q4 | Q1 | Q2-Q3 | Q4 | Q1 | Q2-Q3 | Q4 | |
| No. | 5268 | 8818 | 4474 | 5159 | 9293 | 4108 | 5123 | 9372 | 4065 |
| Provegetarian score, range | 29-53 | 54-61 | 62-80 | 33-52 | 53-62 | 63-87 | 31-52 | 53-62 | 63-86 |
| Age, y | 35.5±11.7 | 37.7±11.9 | 39.8±12.1 | 33.3±10.2 | 37.8±11.8 | 42.5±12.5 | 40.8±12.5 | 37.2±11.8 | 34.3±10.9 |
| Female sex, % | 61.3 | 62.6 | 60.9 | 61.6 | 62.4 | 60.8 | 61.7 | 62.2 | 61.9 |
| Married, % | 44.7 | 50.0 | 53.4 | 38.9 | 51.0 | 58.5 | 55.9 | 49.0 | 41.8 |
| Years of university education | 5.0±1.5 | 5.0±1.5 | 5.2±1.6 | 5.0±1.5 | 5.0±1.5 | 5.1±1.6 | 5.1±1.5 | 5.1±1.5 | 5.0±1.4 |
| BMI, kg/m2 | 23.3±3.5 | 23.5±3.5 | 23.5±3.5 | 23.1±3.4 | 23.5±3.5 | 23.6±3.5 | 23.9±3.6 | 23.4±3.5 | 23.0±3.4 |
| Physical activity, METs-h/wk | 21.5±23.5 | 21.8±22.6 | 23.5±24.1 | 20.0±22.0 | 21.6±22.3 | 26.1±26.3 | 23.9±23.9 | 22.0±23.4 | 20.3±21.9 |
| Smoking, % | |||||||||
| Never smoker | 52.8 | 49.8 | 46.8 | 56.1 | 49.1 | 44.2 | 45.3 | 50.1 | 55.5 |
| Current smoker | 26.3 | 25.2 | 25.0 | 27.4 | 25.9 | 22.3 | 23.5 | 26.2 | 26.4 |
| Former smoker | 20.9 | 24.9 | 28.2 | 16.6 | 25.1 | 33.5 | 31.2 | 23.8 | 18.1 |
| Television watching, h/d | 4.7±2.8 | 4.6±2.8 | 4.7±2.8 | 4.6±2.8 | 4.6±2.8 | 4.7±2.9 | 4.7±2.8 | 4.6±2.8 | 4.7±2.8 |
| Snacking between meals, % | 38.6 | 34.1 | 30.6 | 42.1 | 33.1 | 28.2 | 27.8 | 34.1 | 43.9 |
| Following a special diet, % | 5.9 | 7.9 | 9.4 | 4.6 | 7.5 | 11.9 | 12.5 | 6.8 | 3.4 |
| Alcohol intake, g/d | 7.1±11.7 | 6.6±9.8 | 6.4±8.6 | 5.7±8.4 | 6.7±10.0 | 7.9±12.1 | 6.4±9.1 | 6.8±9.9 | 7.0±11.8 |
| Adherence to Mediterranean diet, Trichopoulou, % | |||||||||
| Low (0-3) | 54.0 | 33.9 | 15.7 | 63.0 | 31.6 | 8.4 | 19.4 | 37.1 | 50.9 |
| Medium (4-5) | 35.8 | 41.4 | 38.3 | 31.0 | 45.1 | 35.6 | 40.3 | 39.9 | 35.7 |
| High (6-9) | 10.2 | 24.7 | 46.0 | 6.1 | 23.2 | 56.0 | 40.3 | 23.1 | 13.5 |
| Supplement use, % | 18.8 | 18.9 | 19.8 | 17.9 | 18.9 | 21.0 | 20.2 | 18.6 | 18.9 |
| Use of cardiovascular drugs, % | 2.1 | 2.6 | 3.0 | 1.7 | 2.5 | 3.8 | 3.3 | 2.4 | 2.0 |
| Hypertension at baseline, % | 9.0 | 10.2 | 10.5 | 7.0 | 10.2 | 12.9 | 12.6 | 9.5 | 7.6 |
| Hypercholeste-rolemia at baseline, % | 12.7 | 16.7 | 19.8 | 11.8 | 16.3 | 22.1 | 20.0 | 15.8 | 13 |
| Diabetes at baseline, % | 1.5 | 1.6 | 1.7 | 0.9 | 1.6 | 2.5 | 3.0 | 1.3 | 0.6 |
| Cancer at baseline, % | 2.2 | 2.6 | 2.8 | 1.6 | 2.6 | 3.7 | 3.4 | 2.3 | 1.9 |
| Family history of CVD, % | 12.9 | 13.0 | 15.2 | 11.1 | 13.7 | 16.1 | 15.3 | 13.1 | 12.2 |
BMI, body mass index; CVD, cardiovascular disease; hPVG, healthful provegetarian; METs, metabolic equivalent of task; PVG, provegetarian; Q, quartile; uPVG, unhealthful provegetarian.
Unless otherwise indicated, the data are expressed as mean±standard deviation.
The associations between each PVG food pattern and cardiovascular risk in the SUN cohort study are described in table 2 and figure 1. Participants in the highest quartile of the PVG food pattern had a significantly lower risk of confirmed CVD during follow-up than those in the lowest quartile in the fully adjusted model (HRQ4vsQ1, 0.63; 95%CI, 0.42-0.94; P for trend=.048). Similarly, the fully adjusted model showed that participants with high adherence to the hPVG food pattern had lower CVD risk compared with those with low adherence (HRQ4vsQ1, 0.60; 95%CI, 0.40-0.90); P for trend=.040]. In contrast, individuals in the highest quartile of the uPVG food pattern had a 76% higher CVD risk after multivariable adjustment compared with those in the lowest quartile (HRQ4vsQ1, 1.76; 95%CI, 1.17-2.64) with a significant dose-response relationship (P for trend=.006). In repeated measurements, the associations between PVG, hPVG, and uPVG food patterns and CVD risk remained significant (PVG: HRQ4vsQ1, 0.66; 95%CI,0.44-0.99; P for trend=.100; hPVG: HRQ4vsQ1, 0.59; 95%CI, 0.39-0.89; P for trend=.041; uPVG: HRQ4vsQ1, 1.74; 95%CI, 1.16-2.61; P for trend=.008).
Cox proportional hazards ratios and 95% confidence intervals for cardiovascular disease according to quartiles of each provegetarian food pattern
| PVG | Q1 | Q2 | Q3 | Q4 | P for trend |
|---|---|---|---|---|---|
| No. | 5268 | 4374 | 4444 | 4474 | |
| Cases/person-y | 64/76869 | 45/63819 | 65/64175 | 53/63682 | |
| Sex- and age-adjusted | 1.00(ref.) | 0.71(0.49-1.03) | 0.88(0.62-1.25) | 0.66(0.45-0.95) | .048 |
| Multivariable-adjusted model | 1.00(ref.) | 0.66(0.44-0.98) | 0.86(0.59-1.25) | 0.63(0.42-0.94) | .048 |
| Repeated measurements of diet | 1.00(ref.) | 0.66(0.44-0.99) | 0.89(0.61-1.30) | 0.66(0.44-0.99) | .100 |
| hPVG | Q1 | Q2 | Q3 | Q4 | P for trend |
| No. | 5159 | 4912 | 4381 | 4108 | |
| Cases/person-y | 45/75571 | 54/71958 | 66/63414 | 62/57603 | |
| Sex- and age-adjusted | 1.00(ref.) | 0.73(0.49-1.10) | 0.90(0.61-1.32) | 0.63(0.42-0.94) | .062 |
| Multivariable-adjusted model | 1.00(ref.) | 0.71(0.48-1.07) | 0.86(0.59-1.29) | 0.60(0.40-0.90) | .040 |
| Repeated measurements of diet | 1.00(ref.) | 0.66(0.44-1.01) | 0.91(0.62-1.34) | 0.59(0.39-0.89) | .041 |
| uPVG | Q1 | Q2 | Q3 | Q4 | P for trend |
| No. | 5123 | 4992 | 4380 | 4065 | |
| Cases/person-y | 75/71516 | 51/72512 | 54/64098 | 47/60419 | |
| Sex- and age-adjusted | 1.00 (ref.) | 0.94(0.65-1.35) | 1.23(0.86-1.76) | 1.68(1.15-2.46) | .009 |
| Multivariable-adjusted model | 1.00 (ref.) | 1.02(0.70-1.50) | 1.30(0.90-1.90) | 1.76(1.17-2.64) | .006 |
| Repeated measurements of diet | 1.00 (ref.) | 1.00(0.69-1.47) | 1.30(0.91-1.88) | 1.74(1.16-2.61) | .008 |
hPVG, healthful provegetarian; PVG, provegetarian; Q, quartile; uPVG, unhealthful provegetarian.
All models were stratified by age groups (10-year periods) and recruitment period (6-year periods). Multivariable-adjusted model adjusted for age, sex, body mass index (kg/m2, linear and quadratic terms, continuous), marital status (5 categories), years of university education (continuous), physical activity (continuous), television watching (continuous), total energy intake (continuous), cumulative smoking (packs-y, 4 categories), smoking status (3 categories), alcohol intake (g/d, 3 categories), following a special diet (dichotomous) and snacking (dichotomous), prevalent cancer (dichotomous), prevalent diabetes (dichotomous), prevalent hypertension (dichotomous), prevalent hypercholesterolemia (dichotomous), prevalent hypertriglyceridemia (dichotomous), use of cardiovascular drugs, and family history of CVD (dichotomous).
Cardiovascular disease incidence according to quartiles of each provegetarian food pattern. Adjusted for age, sex, body mass index, marital status, years of university education, physical activity, television watching, energy intake, cumulative smoking, smoking status, alcohol intake, special diet and snacking, cancer, diabetes, hypertension, hypercholesterolemia, hypertriglyceridemia, cardiovascular drugs, and family history of CVD. 95%CI, 95% confidence interval; CVD, cardiovascular disease; hPVG, healthful provegetarian; PVG, provegetarian; Q, quartil; uPVG, unhealthful provegetarian.
We did not find a statistically significant interaction between PVG dietary patterns and sex, although the effect was more apparent among men than women. We found an interaction between the uPVG food pattern and BMI (P for interaction=.013). The nominal P value was also significant for the effect modification by physical activity (P for interaction=.040). However, these interactions did not remain statistically significant after correction for multiple comparisons using the Bonferroni method (corrected P-values: BMI=0.065; physical activity=0.200) (figure 2, figure 3 and figure 4).
Subgroup analyses for the association between provegetarian food pattern and cardiovascular disease incidence. Adjusted for age, sex, body mass index, marital status, years of university, physical activity, television watching, energy intake, cumulative smoking, smoking status, alcohol intake, special diet and snacking, cancer, diabetes, hypertension, hypercholesterolemia, hypertriglyceridemia, cardiovascular drugs, and family history of CVD. 95%CI, 95% confidence interval; BMI, body mass index; CVD, cardiovascular disease; HR, hazard ratio; METs, metabolic equivalent of task; PVG, provegetarian.
Subgroup analyses for the association between healthful provegetarian food pattern and cardiovascular disease incidence. Adjusted for age, sex, BMI, marital status, years of university, physical activity, television watching, energy intake, cumulative smoking, smoking status, alcohol intake, special diet and snacking, cancer, diabetes, hypertension, hypercholesterolemia, hypertriglyceridemia, cardiovascular drugs, and family history of CVD. 95%CI, 95% confidence interval; BMI, body mass index; CVD, cardiovascular disease; HR, hazard ratio; hPVG, healthful provegetarian; METs, metabolic equivalent of task.
Subgroup analyses for the association between unhealthful provegetarian food pattern and cardiovascular disease incidence. Adjusted for age, sex, body mass index, marital status, years of university, physical activity, television watching, energy intake, cumulative smoking, smoking status, alcohol intake, special diet and snacking, cancer, diabetes, hypertension, hypercholesterolemia, hypertriglyceridemia, cardiovascular drugs, and family history of CVD. 95%CI, 95% confidence interval; CVD, cardiovascular disease; HR, hazard ratio; METs, metabolic equivalent of task; uPVG, unhealthful provegetarian.
We observed similar associations between PVG, hPVG, and uPVG food patterns and CVD after excluding individuals with diabetes, cancer, family history of CVD at baseline, and CVD cases occurring within the first 2 years of follow-up. The findings remained consistent after additional adjustments for margarine intake, vegetable fats and oils, sodium, supplement use, and depression. However, no significant association was observed between the PVG and hPVG food patterns and CVD risk when we excluded individuals with an energy intake of <500 or >3500kcal/d in women, and <800 or> 4000kcal/d in men, or when we excluded those on special diets at baseline. When we excluded CVD deaths, the PVG pattern showed no association with CVD risk. However, the results for hPVG and uPVG remained consistent with the main analyses, showing increased association magnitudes (table 3).
Sensitivity analyses for the association between each provegetarian food pattern and cardiovascular disease in different scenarios (highest vs lowest quartile)
| Analyses | No. | Incident | PVG | hPVG | uPVG |
|---|---|---|---|---|---|
| HR(95%CI) | HR(95%CI) | HR(95%CI) | |||
| Overall | 18 560 | 227 | 0.63(0.42-0.94) | 0.60(0.40-0.90) | 1.76(1.17-2.64) |
| Excluding participants with | |||||
| Implausible total energy intake (*) | 18 611 | 244 | 0.77(0.53-1.11) | 0.68(0.46-1.01) | 1.78(1.22-2.58) |
| Special diet at baseline | 17 141 | 205 | 0.66(0.43-1.01) | 0.66(0.43-1.01) | 1.87(1.22-2.86) |
| Diabetes at baseline | 18 261 | 209 | 0.62(0.41-0.93) | 0.57(0.37-0.89) | 2.06(1.35-3.14) |
| Cancer at baseline | 18 089 | 219 | 0.62(0.41-0.93) | 0.61(0.40-0.92) | 1.71(1.14-2.56) |
| Family history of cardiovascular disease | 16 054 | 175 | 0.59(0.38-0.93) | 0.56(0.36-0.88) | 1.89(1.21-2.95) |
| CVD deaths | 18 446 | 113 | 0.72(0.43-1.21) | 0.49(0.28-0.84) | 2.20(1.27-3.81) |
| Early cases(first 2 y) | 18 408 | 200 | 0.71(0.46-1.09) | 0.56(0.36-0.87) | 1.86(1.21-2.87) |
| No FFQ completed at the 10-year follow-up | 6809 | 58 | 0.88(0.42-1.85) | 0.51(0.23-1.14) | 2.26(1.00-5.08) |
| Additionally adjusted for | |||||
| Margarine | 18 560 | 227 | 0.63(0.43-0.94) | 0.60(0.40-0.90) | 1.75(1.17-2.62) |
| Vegetable fats and oils | 18 560 | 227 | 0.63(0.42-0.94) | 0.60(0.40-0.90) | 1.76(1.17-2.63) |
| Sodium intake | 18 560 | 227 | 0.64(0.43-0.97) | 0.62(0.40-0.94) | 1.77(1.18-2.64) |
| Supplement intake | 18 560 | 227 | 0.63(0.43-0.94) | 0.60(0.40-0.90) | 1.77(1.19-2.65) |
| Prevalent depression | 18 560 | 227 | 0.63(0.42-0.94) | 0.60(0.40-0.90) | 1.76(1.18-2.63) |
95%CI, 95% confidence interval; CVD, cardiovascular disease; FFQ, Food Frequency Questionnaire; hPVG, healthful provegetarian; HR, hazard ratio; PVG, provegetarian; uPVG, unhealthful provegetarian.
A moderate correlation was observed between the MedDiet and the PVG (r=0.40) and hPVG (r=0.54) food patterns. However, only 35% of the participants were placed in the same quartile for MedDiet and the PVG food pattern, and only 40% were in the same quartile for the MedDiet and the hPVG (table S3).
The multivariable-adjusted results yielded an E-value of 2.553 for the estimate and 1.324 for the 95%CI in the case of the PVG. For the hPVG, the corresponding values were 2.721 for the estimate and 1.462 for the 95%CI. Additionally, the E-value for the uPVG was 2.917 for the estimate and 1.616 for the 95%CI.
DISCUSSIONBoth a general and a healthful PVG food pattern were inversely associated with the incidence of CVD. An unhealthful PVG food pattern was linked to an increased CVD risk (figure 5).
Central illustration. Provegetarian food pattern and cardiovascular disease in the Seguimiento Universidad de Navarra (SUN) cohort. CVD, cardiovascular disease; hPVG, healthful provegetarian; HR, hazard ratio; PVG, provegetarian; Q, quartil; SUN, Seguimiento Universidad de Navarra; uPVG, unhealthful provegetarian.
Participants with higher adherence to the PVG food pattern exhibited a 37% lower CVD risk. Similar findings were observed in other cohort studies.4–6,9–15 However, most cohort studies have been conducted in older participants from non-Mediterranean populations.3–7,9,10,12–15,35–38 To our knowledge, only 1 prospective cohort study has assessed the association between PVG food patterns and CVD risk in Mediterranean populations, although it had a relatively small sample size.11 The ATTICA study conducted similar analyses in the greater metropolitan Athens area, with 2020 participants with a slightly older mean age at baseline than the SUN cohort. Higher adherence to a hPVG food pattern was associated with a 40% lower CVD risk. In contrast, participants in the highest quartile of the uPVG food pattern exhibited a 76% increased CVD risk in the SUN project. Additional studies have also reported an association between the uPVG pattern and CVD.4,5,9,12,15 In the ATTICA study, a positive association was observed between the uPVG pattern and 10-year CVD risk, but this association did not reach statistical significance, possibly due to limited statistical power.11
The association between a hPVG pattern and CVD risk could be explained by several biological mechanisms. Plant-based diets are rich in beneficial nutrients, including polyphenols, plant sterols, plant proteins, and fiber, which have been associated with positive effects on CVD.39–44 Additionally, plant-based diets are low in choline and L-carnitine. Both are metabolized by the gut microbiota into trimethylamine N-oxide, a compound that has been shown to contribute to the progression of atherosclerosis.45,46 Furthermore, plant-based diets contain lower amounts of heme iron, which is found in animal sources such as red meat. Heme iron has been associated with an increased CVD risk.47 The association between the uPVG food pattern and CVD risk can be explained by the high consumption of foods rich in added sugars, refined grains, sodium, and trans fatty acids, all of which contribute to the pathogenesis of CVD.48–52
The MedDiet, emphasizing a greater consumption of plant-based foods, has also been associated with a lower CVD risk.53–55 In our study, there was little overlap between the MedDiet and the hPVG food pattern. This may be due to the differences between the 2 dietary patterns.
Limitations and StrengthsThis study has some limitations. First, we used self-reported dietary information, which might have led to some degree of misclassification. However, the FFQ has been repeatedly validated in Spain.20–22 Furthermore, the high educational level of our participants enhances the quality of the self-reported data.56 Second, since the participants are university graduates, the generalizability of the findings to the general population may be limited. However, restricting the study to highly educated participants helped to minimize potential confounding due to educational level.57 Third, due to the inherent limitations of observational studies, residual confounding cannot be entirely excluded; the inability to control for all confounding factors implies that observed associations do not always or necessarily reflect causal relationships. However, the E-values suggest that the observed associations are robust to unmeasured confounding. Fourth, the number of events in the SUN cohort was relatively low, which could affect the estimates obtained and limit the statistical power to detect significant associations in certain subgroups. Additionally, after adjustment for confounders, the association between PVG and CVD was significant in Q2 but not in Q3, potentially reflecting greater misclassification of exposure in the intermediate quartiles.
Among the strengths of the present study are the large sample size, the prospective design, the extensive follow-up, the possibility to adjust for numerous confounding factors, and the evaluation of the PVG patterns at baseline and after 10 years of follow-up. Additionally, cardiovascular events were confirmed blind to the exposure using medical records.
CONCLUSIONSIn a Mediterranean population of university graduates who typically exhibit healthier dietary habits than the general population, greater adherence to a hPVG food pattern was associated with a reduced CVD risk. In contrast, higher adherence to an uPVG food pattern was associated with an increased CVD risk.
FUNDINGThis research was funded by the Spanish Government-Instituto de Salud Carlos III, the European Regional Development Fund (FEDER) (RD 06/0045, CIBEROBN, Grants PI10/02658, PI10/02293, PI13/00615, PI14/01668, PI14/01798, PI14/01764, PI17/01795, PI21/00564, PI23/01132, PI24/01723 and G03/140), the Navarre Regional Government (27/2011, 45/2011, 122/2014, and 19/2023), the National Plan on Drugs (2020/021), and the University of Navarre.
ETHICAL CONSIDERATIONSThis study was conducted in accordance with the principles of the Declaration of Helsinki. The voluntary completion of the baseline questionnaire was viewed as an indication of informed consent. The Research Ethics Committee at the University of Navarra approved this approach for obtaining informed consent from participants, and the Human Research Ethical Committee at the University of Navarre granted approval (2001/30).
SAGER guidelines have been followed. The SUN Project is composed of approximately 60% women; therefore, women are extensively represented. In addition, all multivariable analyses have been adjusted for sex, and we conducted stratified analyses assessing different results among men and women.
STATEMENT ON THE USE OF ARTIFICIAL INTELLIGENCEAll data were collected and analyzed without the use of any artificial intelligence tools.
AUTHORS’ CONTRIBUTIONSConceptualization: A. Martínez-Tabar, M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Methodology: A. Martínez-Tabar, M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Software: M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Validation: M.Á. Martínez-González, M. Ruiz-Canela, M. Bes-Rastrollo, and J. Díaz-Gutiérrez. Formal analysis: A. Martínez-Tabar, M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Investigation: A. Martínez-Tabar, M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Resources: M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Data curation: A. Martínez-Tabar, M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Writing–original draft preparation: A. Martínez-Tabar. Writing–review and editing: V. Bullón-Vela, C. de la Fuente-Arrillaga, C. Sayón-Orea, J. Díaz-Gutiérrez, M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Visualization: M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Supervision: M.Á. Martínez-González, M. Ruiz-Canela, and M. Bes-Rastrollo. Project administration: C. de la Fuente-Arrillaga and M. Bes-Rastrollo. Funding acquisition: M.Á. Martínez-González and M. Bes-Rastrollo.
CONFLICTS OF INTERESTThe authors declare no conflict of interests.
- –
In non-Mediterranean populations, a healthful provegetarian (hPVG) food pattern has been associated with a lower cardiovascular disease (CVD) risk, whereas an unhealthful provegetarian (uPVG) pattern has been associated with a higher risk.
- –
There is a lack of evidence regarding the effect of provegetarian (PVG) food patterns on CVD risk in Mediterranean populations.
- –
This is the first large, prospective study to examine the effect of PVG food patterns on CVD in a relatively young Mediterranean population.
- –
Higher adherence to a PVG and hPVG food pattern was associated with a lower CVD risk. Greater adherence to an unhealthful PVG food pattern was associated with a higher CVD risk.
The authors express their appreciation to the SUN participants for their enthusiastic and continued collaboration with the project, as well as to the SUN staff and researchers for their exceptional work.
Supplementary data associated with this article can be found in the online version, at https://doi.org/10.1016/j.rec.2025.11.008.