Highlights
-
•
High cumulative EVOO intake (>40 g/d) reduced major cardiovascular disease risk.
-
•
EVOO lowered the risk of stroke, peripheral artery disease, and atrial fibrillation.
-
•
Common olive oil showed weaker associations compared to EVOO.
-
•
Intake was evaluated with yearly repeated measures over 4.7 years of follow-up.
ABSTRACT
Background
Olive oil, increasingly consumed in the U.S., has been inversely associated with cardiovascular disease (CVD) risk. However, previous studies did not assess a broad spectrum of CVD outcomes, incorporated repeated annual dietary assessments, or distinguished between polyphenol-rich EVOO and common olive oil (COO), which lacks these compounds.
Methods
We assessed 7102 high-risk participants from the PREDIMED trial (57.5% women; aged 55-80 years), all free of CVD at baseline. Olive oil consumption was assessed annually, and cumulative average intakes of EVOO and COO were calculated. The primary outcome was a composite of myocardial infarction, stroke, peripheral arterial disease, heart failure, atrial fibrillation, or cardiovascular death, whichever occurred first. Individual outcomes were also evaluated. Time-dependent Cox models were adjusted for major confounders, including trial intervention arm.
Results
Over a median follow-up of 4.7 years, 621 participants experienced at least one CVD event. Participants in the highest tertile of cumulative EVOO intake (mean: 49.2 g/d) had a 25% lower risk of the composite outcome (HR: 0.75; 95% CI: 0.60-0.94), with significant reductions in several individual CVD outcomes. In the decile analysis, the highest (mean: 60.9 g/d) versus lowest decile had a 48% lower risk (HR: 0.52; 95% CI: 0.35 to 0.79). COO consumption was not significantly associated with CVD risk when mutually adjusted for EVOO (HR per 10 g/d : 0.93; 95% CI: 0.87-1.00).
Conclusions
High consumption of EVOO is associated with a substantial reduction in the risk of an outcome-wide composite of CVD events among high-risk individuals. In contrast, COO, which lacks polyphenols, showed weaker associations, highlighting the importance of differentiating olive oil types in CVD prevention strategies.
Trial Registration
This trial was registered in the ISRCTN registry (ISRCTN 35739639): https://www.isrctn.com/ISRCTN35739639 .
Graphical abstract
The traditional Mediterranean diet (MedDiet) has long been recognized as a model for healthy eating and cardiovascular disease (CVD) prevention, supported by a robust body of epidemiological and clinical trial evidence, particularly over the past three decades. A central feature of this dietary pattern is the emphasis on fat quality, with olive oil as the primary source of fat
Olive oil consumption has risen sharply outside Mediterranean countries, including the United States, where intake increased from about 28,000 metric tons in the 1970s to over 400,000 metric tons in the 2020s Extra-virgin olive oil (EVOO) is obtained through mechanical cold pressing of olives—without chemical solvents or heat—thereby preserving a wide range of bioactive phenolic compounds. These polyphenols, such as oleocanthal and hydroxytyrosol, are efficiently absorbed and may contribute to EVOO’s health benefits beyond its high monounsaturated fat content. , In contrast, common olive oil (COO) typically consists of refined olive oil (≥80%) blended with a small proportion of virgin olive oil (≤20%). The refining process—often involving heat, chemical solvents, and deodorization—removes nearly all polyphenols and other bioactive compounds, leaving primarily monounsaturated fats.
Although their fat composition is similar, EVOO and COO differ markedly in their antioxidant and anti-inflammatory properties, which may translate into different cardiovascular effects. , As a result, EVOO has gained increasing recognition for its potential to reduce the risk of CVD—including myocardial infarction (MI), stroke, and cardiovascular death—in individuals at high CVD risk Despite these findings, important gaps remain. To date, no clinical study has comprehensively evaluated the associations between different olive oil types and a broad spectrum of major CVD outcomes—including peripheral artery disease (PAD), heart failure (HF), and atrial fibrillation (AF). Moreover, previous studies have largely relied on baseline dietary data, without accounting for dietary changes over time.
For instance, Guasch-Ferré et al. analyzed only baseline olive oil consumption and a limited composite CVD endpoint—MI, stroke, or CVD death—in the PREDIMED trial, which recorded 277 total events. Their analysis did not include other CVD outcomes such as PAD, HF, or AF, nor did it account for cumulative intake throughout follow-up. Other cohort studies using repeated dietary assessments did not distinguish between EVOO and COO and were conducted in U.S. populations with substantially lower olive oil consumption. , Furthermore, dietary data were collected every four years rather than annually. In contrast, yearly repeated dietary assessments may provide a more accurate and time-sensitive characterization of diet as a dynamic exposure
Observational data from the PREDIMED study provide a valuable opportunity to address these limitations. While intention-to-treat effects of the MedDiet interventions on CVD incidence have been previously reported, ,,, those analyses did not account for actual adherence to dietary recommendations, particularly regarding EVOO consumption. Importantly, unlike most previous studies, PREDIMED allows for separate analyses of EVOO and COO, enabling a more nuanced assessment of their respective associations with CVD outcomes.
The aim of this study was to provide a robust evaluation of the associations between each olive oil type—EVOO and COO—and a wide range of adjudicated CVD outcomes, using annually repeated validated measures of intake in the context of a well-known randomized primary prevention trial (PREDIMED). By applying an outcome-wide analytical approach, we assessed both a composite CVD endpoint and its individual components: MI, stroke, PAD, HF, AF, and cardiovascular death.
Methods
Data source and ethical considerations
The design and primary results of the PREDIMED study (ISRCTN: 35739639) have been published The Institutional Review Boards of all the recruitment centers approved the overall PREDIMED trial design according to the ethical guidelines of the Declaration of Helsinki. All participants gave written informed consent. The present study adhered to most of the items in the SAGER (Sex and Gender Equity in Research) guidelines checklist. Our sample was sex-balanced, with 57.5% women. Sex differences were specifically analyzed and considered in the data analysis and interpretation where relevant. The completed SAGER checklist is provided in the Supplementary Material.
Study design and population
The PREDIMED study was a multicenter randomized primary CVD prevention trial conducted in Spain in 7,447 participants between 55 and 80 years of age, without prevalent CVD at the time of recruitment but at high CVD risk. , Participants were randomized to one of the following three dietary interventions: a Mediterranean diet supplemented with extra-virgin olive oil (MedDiet with EVOO), a Mediterranean diet supplemented with nuts (MedDiet with nuts), or a control diet consisting of advice to follow a low-fat diet. Recruitment began in June 2003 and concluded in June 2009.
This observational analysis was conducted as a prospective cohort study assessing PREDIMED trial participants. In subtype-specific analyses, participants who developed CVD outcomes other than the one being analyzed, were censored at that point in time. Additionally, AF was not consistently assessed as a relevant endpoint at one of the eleven research centers, resulting in the exclusion of 674 participants from our AF analysis.
Ascertainment of cardiovascular disease
We included only CVD events that were reviewed and blindly adjudicated by the Clinical Event Committee until December 1, 2010. Committee members were blinded to both treatment allocation and participants’ dietary information. CVD outcomes included MI, stroke, PAD, HF, AF, and cardiovascular mortality. Detailed diagnostic criteria are provided in the Supplementary Material.
Dietary evaluation
A 137-item validated food frequency questionnaire (FFQ) was administered yearly in face-to-face interviews by trained dietitians. The FFQ included questions about “olive oil,” “pomace olive oil,” and “extra-virgin olive oil,” with intake measured in tablespoons (10 g each). Participants reported their consumption frequency using nine categories: never, 1-3 times per month, weekly (1, 2-4, 5-6 times), or daily (1, 2-3, 4-6, > 6 times). Responses were converted into grams per day. COO intake was determined by summing the first two items, while the third item was used to assess EVOO intake. Information from the FFQ was also used to estimate total energy intake. The FFQ’s reproducibility and validity have been previously confirmed Additionally, olive oil consumption as measured through the FFQ has been associated with a distinct metabolomic profile in the PREDIMED study. This profile has been linked to a reduced risk of incident CVD More recently, dietary patterns, with particular attention to refined carbohydrates and their overall quality, derived from the same FFQ in the PREDIMED-Plus study, have also been associated with favorable changes in CVD risk factors
Refined carbohydrate intake was estimated from FFQ items corresponding to white bread, refined breakfast cereals, white rice, refined pasta, pizza, biscuits and cookies (plain or chocolate), cakes, muffins, donuts, croissants, pastries, sponge cakes, chocolate bars, churros, savory snacks, custards, ice cream, canned fruits in syrup, sugar-sweetened beverages, packaged fruit juices, nougat, marzipan, table sugar, jams, and honey. For each item, carbohydrate content was calculated using standard portion sizes and nutrient composition values and then summed to obtain total refined carbohydrate intake (g/d).
Adherence to the traditional MedDiet was also assessed annually using the validated 14-item Mediterranean Adherence Screener (MEDAS). To evaluate dietary patterns independent of olive oil consumption, two olive oil-related items were excluded, resulting in a 12-point MEDAS score. This modified version of the MEDAS was used as a covariate in statistical models to adjust for overall adherence to the MedDiet
Anthropometric measures and other covariates
Blood pressure and anthropometric measurements, including weight, height, and waist circumference, were recorded in accordance with the study protocol by trained personnel. Other covariates collected in the PREDIMED trial have been previously described
Statistical analyses
The present observational analysis was conducted as a prospective cohort study. Participants with extreme reported total energy intake at baseline—defined as exceeding 4,000 or falling below 800 kcal per day for men and exceeding 3,500 or falling below 500 kcal per day for women—were excluded from the current analysis.
We used multivariable time-dependent Cox models to estimate hazard ratios (HR) for developing total CVD and each of its individual components. Person-years of follow-up were calculated from the baseline visit date until the occurrence of a CVD event, death, or the end of follow-up. Olive oil consumption was adjusted for total energy intake using the residual model In the main models, the relevant exposure for each type of olive oil was the cumulative mean of intake, equivalent to the average values from each year up to the last follow-up visit. In an ancillary analysis, we used updated intake data, reflecting only the most recent measurement at each time point, without averaging previous values.
To control for potential confounding factors, a multivariable model was adjusted for age, educational level, body mass index (BMI), waist-to-height ratio, physical activity level at baseline, cumulative total energy intake, cumulative 12-point MEDAS score, smoking status, baseline prevalence of type 2 diabetes, dyslipidemia, hypertension, and family history of early-onset CHD. A second model was additionally adjusted for the use of statins, other lipid-lowering agents, angiotensin-converting enzyme (ACE) inhibitors or angiotensin receptor blockers (ARBs), diuretics, other antihypertensive agents, acetylsalicylic acid, insulin, and oral hypoglycemic agents. Finally, a third model further adjusted for propensity scores (based on 30 baseline variables estimating the probability of assignment to each intervention group) and by the randomized intervention group. All models were stratified by sex and recruitment center and used robust variance estimators, to account for intra-cluster correlations.
Hazard ratios were computed across three categories according to sex-specific tertiles of each type of olive oil intake, with the first tertile as the reference category. To examine linear trends, we conducted a Wald test for linearity by considering the median intake within each tertile as a continuous variable. An additional HR was computed for each 10-gram per day increment in olive oil intake. We also conducted isocaloric substitution analyses to estimate the effect of replacing 100 kcal/d of COO with EVOO, and 100 kcal/d of refined carbohydrates with EVOO.
We conducted subgroup analyses using predefined subgroups, including sex, age, intervention group, BMI, and smoking status. Potential interactions between these stratification factors were assessed through likelihood ratio tests.
The cumulative incidence of CVD was plotted using Nelson-Aalen incidence curves, based on the joint classification of participants by average EVOO intake (< 40 g/d or ≥ 40 g/d) from baseline to the 7th year of follow-up, and by the randomized intervention arm of the trial (MedDiet or control group). These graphs were adjusted using inverse probability weighting, employing the same covariables as in the Cox regression models.
Results were presented with 95% confidence intervals (CI), and statistical significance was determined by a two-tailed p-value less than 0.05. All analyses were performed using STATA statistical software version 16 (StataCorp. 2019. Stata Statistical Software: Release 16, College Station, TX: StataCorp LLC).
Results
Description of participants
Following the exclusion of individuals with total energy intake out of predefined limits ( n = 153), incomplete baseline dietary data ( n = 78), and pre-existing symptoms of HF, AF, or intermittent claudication ( n = 114), 7,102 participants remained for analysis (see Supplementary Figure 1 for the flowchart).
The mean age of the participants was 67 years and 57.5% of them were women. Baseline characteristics of participants, categorized by tertiles of average EVOO intake during follow-up, are detailed in Table 1 . A higher average intake of EVOO during follow-up was correlated with younger age, a lower percentage of women, higher level of education, lower BMI, lower waist-to-height ratio, lower percentage of obesity, higher physical activity at baseline, higher adherence to the MedDiet (based on the 14-point MEDAS score) at baseline, a higher percentage of former smokers, and a lower percentage of dyslipidemia.
Table 1
Baseline characteristics of participants according to tertile of average EVOO intake during follow-up
| Tertile of average EVOO intake during follow-up | T1 | T2 | T3 |
|---|---|---|---|
| N | 2,368 | 2,367 | 2,367 |
| Mean follow-up EVOO intake (SD), g/d | 8.5 (8.0) | 33.0 (5.5) | 50.9 (6.6) |
| Female sex % | 60.60 % | 59.5 % | 52.9 % |
| Mean age (SD), years | 67.5 (6.3) | 66.9 (6.1) | 66.6 (6.1) |
| Mean education level (SD), years | 3.7 (2.0) | 3.9 (2.2) | 4.2 (2.4) |
| Marital status, % married | 73.9 % | 76.1 % | 78.9 % |
| Mean BMI (SD), kg/m² | 30.1 (4.0) | 29.9 (3.9) | 29.8 (3.7) |
| Mean waist-to-height ratio (SD) | 0.64 (0.07) | 0.63 (0.07) | 0.62 (0.06) |
| Mean leisure-time physical activity level (SD), MET min/d | 220 (242) | 220 (235) | 252 (237) |
| Mean total energy intake level (SD), kcal/d | 2,151 (438) | 2,171 (390) | 2,234 (377) |
| Mean MEDAS score (SD) | 9.5 (2.0) | 10.1 (1.9) | 10.7 (1.8) |
| Current smoker % | 14.6 % | 13.0 % | 14.1 % |
| Former smoker % | 21.3 % | 23.7 % | 28.2 % |
| Type 2 diabetes | 50.3 % | 47.7 % | 47.8 % |
| Overweight | 92.9 % | 92.2 % | 92.4 % |
| Obesity | 48.5 % | 46.9 % | 45.2 % |
| Family history of early-onset CHD | 22.1 % | 23.9 % | 21.0 % |
| Dyslipidaemia | 72.8 % | 73.5 % | 70.6 % |
| Hypertension | 82.6 % | 82.3 % | 83.1 % |
| Medication use | |||
| Statins | 40.0 % | 40.8 % | 40.0 % |
| Diuretics | 21.2 % | 20.6 % | 21.8 % |
| Other lipid-lowering agents | 5.1 % | 5.7 % | 5.2 % |
| ACE inhibitors or ARBs | 49.6 % | 47.5 % | 51.0 % |
| Other antihypertensive agents | 28.6 % | 30.3 % | 28.2 % |
| Acetylsalicylic acid | 19.6 % | 19.8 % | 18.5 % |
| Insulin | 5.7 % | 4.9 % | 4.9 % |
| Oral hypoglycaemic agents | 30.9 % | 29.6 % | 28.9 % |
SD: Standard deviation. BMI = Body mass index. MET = Metabolic equivalent. MEDAS = Mediterranean diet adherence score (on a scale of 0 to 14). ACE = Angiotensin-converting enzyme. ARB = Angiotensin receptor blockers.
Detailed changes in cumulative EVOO intake by year throughout the follow-up period for each trial arm are available in Supplemental Figure 2. In the control group, only 20.7% of participants achieved an average EVOO intake equal to or exceeding 40 g per day during the follow-up period. In contrast, in the MedDiet with nuts group, a higher percentage of participants (32.8%) reached this goal of EVOO intake, while in the MedDiet with EVOO group, 83.1% of participants met this threshold.
Outcome data
Over a median follow-up of 4.7 years (interquartile range, 2.8-5.8 years), we identified 621 participants (8.7% of the cohort) who experienced at least one component of the CVD composite outcome—a substantially higher number than the 277 cases (3.9% of the cohort) included in prior PREDIMED assessments focused solely on MI, stroke, and cardiovascular death In our current analysis, we included 103 adjudicated cases of myocardial infarction (MI), 135 of stroke, 87 of peripheral arterial disease (PAD), 90 of heart failure (HF), 245 of atrial fibrillation (AF), and 81 cardiovascular-related mortality events (the total exceeds 621 because some participants experienced more than one event).
Crude incidence rates of the composite CVD endpoint per 100 person-years were 2.5 for the lowest tertile of cumulative EVOO intake, 2.1 for the second tertile, and 1.5 for the third tertile ( Table 2 ). These rates showed, therefore, a consistent decreasing trend as cumulative EVOO intake increased. In the multivariable-adjusted model, which accounted for total energy intake, participants in the highest tertile of cumulative EVOO intake (mean: 49.2 g/d) had a 32% relative reduction in the risk of CVD events as compared to those in the lowest tertile (hazard ratio [HR]: 0.68; 95% confidence interval [CI]: 0.55 to 0.85; P for trend =.001). Further adjustment for the randomized intervention group slightly attenuated these results, resulting in an HR of 0.75 (95% CI: 0.60-0.94; P for trend =.02) for the highest versus the lowest tertile. When cumulative EVOO intake was divided into deciles instead of tertiles, a linear dose-response relationship emerged between EVOO intake and CVD risk, with no apparent threshold for the beneficial effects of EVOO consumption ( Table 3 ). Comparing the highest (mean: 60.9 g/d) versus lowest deciles of cumulative EVOO intake, the HR was 0.52 (95% CI: 0.35-0.79) in the multivariable model, which included adjustment for the randomized intervention group. Ancillary analyses using updated intake data (i.e., considering only the most recent measurement at each time point) showed nearly identical results as compared to cumulative intake (see Supplementary Table 1 ).
Table 2
Time-dependent Cox regression models assessing the risk of overall CVD by energy-adjusted tertiles of cumulative EVOO intake (g/d) during follow-up and as a continuous variable (per 10 g/d)
| Tertiles for cumulative EVOO intake | Per 10 g/d increase in EVOO intake | ||||
|---|---|---|---|---|---|
| T1 | T2 | T3 | |||
| N | 2368 | 2368 | 2366 | ||
| Mean cumulative EVOO intake (g/d) | 6.3 (7.3) | 27.5 (5.1) | 49.2 (9.4) | ||
| N of cases | 261 | 209 | 151 | ||
| Person-years | 10256 | 9822 | 9910 | ||
| Incidence rate per 100 person-years (95% CI) | 2.5 (2.3- 2.9) | 2.1 (1.9- 2.4) | 1.5 (1.3- 1.8) | ||
| Models | p for trend | ||||
| Multivariate adjusted A * | 1 (ref.) | 0.93 (0.77- 1.13) | 0.68 (0.55- 0.85) | 0.001 | 0.91 (0.87- 0.95) |
| Multivariate adjusted B † | 1 (ref.) | 0.92 (0.76- 1.12) | 0.68 (0.55- 0.85) | 0.001 | 0.91 (0.87- 0.95) |
| Multivariate adjusted C § | 1 (ref.) | 0.99 (0.81- 1.21) | 0.75 (0.60- 0.94) | 0.02 | 0.93 (0.89- 0.97) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
