We carried out a systematic review and meta-analysis following the recommendations established by the PRISMA statement (Preferred Reporting Items for Systematic Reviews and Meta-Analyzes) (Moher et al., 2009). A systematic review protocol was developed, not registered a priori, and not modified after the beginning of the review process.

We used the following inclusion criteria: a) participants: adult population with at least 18 years of age without previous cardiovascular events; b) intervention: diet enriched with olive oil for at least 12 weeks vs. diet enriched with other fat; c) outcome measures: SBP or DBP; d) design of included studies: randomized controlled trials (RCTs). We have included both healthy individuals with cardiovascular risk factors or with other non-cardiovascular pathology.

We exclude studies in which the intervention was carried out in cardiovascular patients because they experience a profound change in their lifestyle, including diet, not comparable with healthy subjects or with other types of diseases. Animal studies were discarded.

The following electronic databases were searched from their inception through April 10, 2018: PubMed, Embase, Cochrane plus, Web of Science, Ovid, Scopus, VHL, TDR. The language of publication was not restricted.

The search strategy was carried out through the combination of keywords related to olive oil intake (olive oil), and the different expected results: hypertension (arterial pressure, hypertension, blood pressure, DBP, SBP), and the design of adequate studies for inclusion (randomized controlled trial, systematic review, meta-analysis). Reference lists of identified original articles were searched manually. Table 1 shows the search strategy for the databases.

We used the Cochrane risk of bias tool to assess risk of bias within the following domains: selection bias, performance bias, detection bias, attrition bias, reporting bias, and other bias (Higgins and Green, 2011). Table 2 shows the studies after the validity analysis.

Two researchers made the selection of articles independently. Discrepancies were discussed with a third review author until consensus was reached. The following data were extracted from each study: authors, year of publication, country of implementation, duration, amount of olive oil intake in the experimental group and fat in the control one, characteristics and number of participants in each intervention applied, SBP and DBP (Table 2).

The outcome parameter was the change in blood pressure regarding baseline values; the mean standard error of the change was computed. Mean differences (MDs) were pooled using a fixed effects model. Heterogeneity between studies results was tested using the Q test. The proportion of heterogeneity was quantified with the I² parameter. I² > 50 % was considered to represent considerable heterogeneity. Metarregression was performed to explore the causes of heterogeneity. We applied Egger’s method to determine the possible existence of publication bias. All data were analyzed using Stata 14 SE program (College Station, TX, USA).

The search identified 872 studies, 687 of which were identified in different databases: 119 in PubMed, 23 in Embase, 32 in the Cochrane Plus Library, 167 in the Web of Science, 35 in Ovid, 232 in Scopus, and 79 in VHL; while 185 were located through the grey literature of TDR and hand searched for references. After the exclusion of duplicates, 691 studies were evaluated. 27 studies were reviewed in full text, and 15 studies were finally included. The reasons for exclusion are detailed in Figure 1.

Fifteen studies were included in the systematic review and meta-analysis (Kristensen et al., 2016, Rozati et al., 2015, Lee et al., 2015, Venturini et al., 2015, Ceriello et al., 2014, Toledo et al., 2013, Singhal et al., 2013, Tapsell et al., 2013, Konstantinidou et al., 2010, Taylor et al., 2006, Rasmussen et al., 2006, Olsen et al., 2000, Ferrara et al., 2000, Prisco et al., 1998, Bonnema et al., 1995). Table 2 shows a summary of the information extracted from the included studies. The designs used by the studies were: double blind parallel RCTs (Kristensen et al., 2016, Lee et al., 2015, Singhal et al., 2013, Taylor et al., 2006, Olsen et al., 2000, Prisco et al., 1998, Bonnema et al., 1995), double-blind, randomized crossover study (Ferrara et al., 2000), and non-blinded RCTs (Rozati et al., 2015, Venturini et al., 2015, Ceriello et al., 2014, Toledo et al., 2013, Tapsell et al., 2013, Konstantinidou et al., 2010, Rasmussen et al., 2006). The follow-up ranged from 12 weeks to 4 years. The countries in which the studies were developed were Spain (Ceriello et al., 2014, Toledo et al., 2013, Konstantinidou et al., 2010), Denmark (Kristensen et al. 2016, Bonnema et al.,1995), Italy (Ferrara et al., 2000, Prisco et al., 1998), United Kingdom (Singhal et al., 2013, Taylor et al., 2006), South Korea (Lee et al., 2015), USA (Rozati et al., 2015), Brazil (Venturini et al., 2015), Australia (Tapsell et al., 2013), and two multinational studies: Denmark, United Kingdom, Sweden, Italy, Netherlands, Belgium, Russia and Norway (Olsen et al., 2000); and Finland, Denmark, Italy, Australia and Sweden (Rasmussen et al., 2006).

The fifteen studies included 6651 participants: 3358 had ingested olive oil and 3293 were fed the control diet. Four studies had more than two arms; in these studies the following branches were discarded for the present review: in the PREDIMED study the branch with Mediterranean diet plus nuts (Toledo et al., 2013), refined olive oil (Konstantinidou et al., 2010), hypocaloric diet plus fish (Tapsell et al., 2013), and pregnant women with fewer weeks of administration (Olsen et al., 2000). Venturini et al., (2015) gives their outcomes as mean and interquartile range and they could not be included in meta-analysis. Men and women were pooled in the analyses.

The participants were adults without cardiovascular events. Some participants had cardiovascular risk factors such as overweight or obesity (Rozati et al., 2015, Tapsell et al., 2013, Taylor et al., 2006), diabetes mellitus (Ceriello et al., 2014), incipient or established diabetic nephropathy (Lee et al., 2015, Bonnema et al., 1995) or more than three cardiovascular risk factors (Toledo et al., 2013), hypertension (Ferrara et al., 2000, Prisco et al., 1998), metabolic syndrome (Venturini et al., 2015) and psoriatic arthritis (Kristensen et al., 2016). Three studies were performed in healthy adults (Singhal et al., 2013, Konstantinidou et al., 2010, Rasmussen et al., 2006). Olsen et al., (2000) recruited twin pregnancies and pregnancy induced hypertension.

The olive oil was administered in capsules of 1-6 g/day (Kristensen et al., 2016, Lee et al., 2015, Singhal et al., 2013, Tapsell et al., 2013, Taylor et al., 2006, Rasmussen et al., 2006, Olsen et al., 2000, Prisco et al., 1998, Bonnema et al., 1995), and in liquid form as extra virgin olive oil (EVOO) from 10 ml to 50 ml/day (Rozati et al., 2015, Venturini et al., 2015, Ceriello et al., 2014, Toledo et al., 2013, Ferrara et al., 2000, Konstantinidou et al., 2010), in the context of different types of diets, such as the American diet (Rozati et al., 2015), or the Mediterranean diet (Konstantinidou et al., 2010, Toledo et al., 2013, Prisco et al., 1998). The studies that supplemented liquid oil provided it as EVOO, while the studies with capsule supplement provided it with olive oil.

The control groups took both capsules, liquid oil, habitual diet and low-fat diet. The controls took capsules with different fatty acids: microalgae and palmitic acid (Singhal et al., 2013), conjugated linoleic acid (Taylor et al., 2006), docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) (Kristensen et al., 2016, Lee et al., 2015, Tapsell et al., 2013, Bonnema et al., 1995, Rasmussen et al., 2006, Olsen et al., 2000, Prisco et al., 1998). In liquid form the next fats were administered to the control groups: corn oil, soybean oil and butter (Rozati et al., 2015), and sunflower oil (Ferrara et al., 2000). Finally, in four studies, no supplement was provided to the control groups: habitual diet (Venturini et al., 2015, Konstantinidou et al., 2010), and low fat diet (Ceriello et al., 2014, Toledo et al., 2013).

The diet administered was adequate to the energy requirements, except in Tapsell et al., (2013) who administered a hypocaloric diet with 1750 (SD 417) kcal and 1600 (SD 355) kcal in the olive oil and control groups, respectively. Physical activity was not restricted in the included studies; they were requested to maintain their regular routine of physical activity. Tapsell et al., (2013), recommended walking for 30-minutes three days a week, with similar results among groups.

Fourteen studies assessed DBP and SBP, and one study only DBP (Olsen et al., 2000). No adverse events were mentioned: only nine participants in the n-3 polyunsaturated fatty acid (PUFA) supplemented group and six participants in the olive oil group reported mild gastrointestinal adverse effects in Kristensen et al., (2016).

We used the Cochrane risk of bias tool (Higgins and Green, 2011) to assess all studies for their risk of bias within the following domains: selection bias, performance bias, detection bias, attrition bias, reporting bias, and other bias. Figure 2 provides an overview of the risk of bias assessment. Two studies had high risk for selective reporting bias (Venturini et al., 2015, Olsen et al., 2000) and one study was classified as high risk for attrition bias (Tapsell et al., 2013). An intention-to-treat analysis was performed. The included studies were considered as having low risk for bias.

Olive oil did not decrease SBP more than other fat, MDs = -0.11, CI 95 % (-0.68, 0.46), I² = 85.1%; p heterogeneity < 0.001. The studies dispensed liquid olive oil or capsules. To identify the effect of the two kinds of supplementation meta-analysis was stratified by this variable (capsules and oil) (Figure 3): no significant effect was observed either for capsules or liquid on SBP.

Olive oil showed a significant decrease in DBP with a fixed effects model of -0.73 mm Hg, 95% CI (-1.07, -0.40); p < 0.001; I² = 84.5%; p heterogeneity < 0.001. This effect was mainly due to liquid oil EVOO: a reduction of -1.44 mm Hg, 95% CI (-1.89, -1.00); p < 0.001; with a high heterogeneity, I² = 86.9%; p heterogeneity < 0.001. The olive oil in capsules had no significant influence on DBP, 0.14 mm Hg, 95% CI (-0.35, 0.64), I² = 76.2%, p = 0.5 (Figure 4). A meta-regression was performed to inquire the causes of heterogeneity. Neither the quantity of grams of olive oil on SBP (p = 0.35) or DBP (p = 0.51), nor the year of publication (p = 0.27 and p = 0.70, respectively), seem to be the cause of the heterogeneity found among the studies selected for the meta-analysis. Among the causes of heterogeneity are the different diets, and the characteristics of the patients for whom the studies were developed.

Egger’s method showed p values of 0.67 and 0.23 for SBP and DBP, respectively.