Network Meta-Analysis of Metabolic Effects of Olive-Oil in Humans Shows the Importance of Olive Oil Consumption With Moderate Polyphenol Levels as Part of the Mediterranean Diet

The beneficial role of olive oil consumption is nowadays widely recognized. However, it is not clear whether its health effects are due to the presence of monounsaturated lipids and/or to the antioxidant fraction of microconstituents present in olive oil. The aim of the present study was to analyze the exact role of olive oil in the modification of metabolic factors (glucose and circulating lipids) and explore the role of its antioxidant polyphenols. In the present work, we have performed a network meta-analysis of 30 human intervention studies, considering direct and indirect interactions and impact of each constituent. Interestingly, we show that the impact of olive oil on glucose, triglycerides, and LDL-cholesterol is mediated through an adherence to the Mediterranean diet, with the only notable effect of olive oil polyphenols being the increase of HDL-cholesterol, and the amelioration of the antioxidant and inflammatory status of the subjects. Additionally, we report for the first time that lower antioxidant polyphenol levels may be sufficient for the beneficial effects of olive oil, while we show that the lipid fraction of olive oil may be responsible for some of its beneficial actions. In all parameters examined the beneficial effect of olive oil was more pronounced in subjects with an established metabolic syndrome or other chronic conditions/diseases. In conclusion, all these findings provide new knowledge that could lead to re-establishment of the role of olive oil in human nutrition.


INTRODUCTION
Impaired glucose and lipid metabolism together with increased blood pressure, that characterize a pro-inflammatory state (1)(2)(3) lead to an increased likelihood of insulin resistance/type 2 diabetes, and atherosclerosis/cardiovascular disease (2), which are the commonest metabolic dysfunctions in humans. These alterations, together with a resulting pre-thrombotic state (3) may result in premature death. Since 1979, after the milestone publication of Keys et al. (4), it was evidenced that the Mediterranean-type of diet resulted in a reduced risk for cardiovascular disease, despite its high (especially mono-unsaturated) lipid content, of which olive oil is the main source. Indeed, a large body of scientific evidence confirmed the benefits of Mediterranean diet and olive oil consumption, on the lipid profile, lipid and DNA oxidation, insulin resistance, and inflammation (5)(6)(7)(8), resulting in a decreased cardiovascular risk.
Olive oil is a functional food that, besides its high content in mono-unsaturated fatty acids, also contains other minor, biologically active, components, such as vitamins, minerals, and polyphenols (9). The quality of olive oil is mainly defined by EU regulations (2568/91/EEC & 1019/2002/EC), which state the requirements for each commercial type: extravirgin olive oil (produced by direct-press or simple, low-speed, centrifugation methods), virgin, common or refined olive oil. However, even extra-virgin olive oil may differ in terms of its microconstituents, related to the method of isolation, microclimate, and cultivation conditions.
Here, we have used published human intervention studies using olive oil, totalizing 7688 subjects, mostly out of the context of the Mediterranean diet and performed a network metaanalysis of metabolic changes in circulating glucose and lipid parameters, in relation to the polyphenol load of the oil. Network meta-analysis reports a final effect of a given intervention, considering both direct and indirect interactions. In addition, we extracted data related to inflammation and oxidation status and analyzed them separately, in a narrative review, as the number of studies was small. We provide novel data on the effect of olive oil and suggest a possible paradigm shift for its use in human nutrition.

MATERIALS AND METHODS
Studies were retrieved from PubMed, Scopus and Google Scholar, using the terms: "(olive oil) AND (intervention study) AND diet AND human" at 20 September 2017. A total of 465 unique records were retrieved (excluding reviews) and scanned for changes of metabolic (glucose, total-, HDL-and LDL cholesterol, triglycerides), markers. The search strategy and details of the excluded studies are presented in Figure 1 (PRISMA flow chart). Details of the retained studies are presented in Table 1. Two or more categories were extracted from each study: (1) Normal (control) diet (CD) or Control Sample (CS); (2) Intervention including Mediterranean Diet with olive oil containing low or high polyphenols; (3) Intervention including olive oil with a low (<60 mg/kg) polyphenol content (LPC); (4) Diet including different categories of olive oil (for example extra-virgin vs. virgin or refined olive oil, etc.). Since our study is a meta-analysis and systematic review of human intervention studies it does not necessitates an ethical committee approval.
From each study, the mean, standard deviation and number of participants per group were extracted and tabulated in Excel. Meta-analysis was performed in Excel, using the free MetaXL V5.3 (www.epigear.com) add-on. Network Meta-Analysis (or indirect method meta-analysis) (53), was performed, using fixed effects. In a network meta-analysis, the effect of two treatments is measured, that each was compared against a similar control group in a meta-analysis. For example, if treatment A and treatment B were directly compared vs. placebo in separate meta-analyses, we can use these two pooled results to get an estimate of the effects of A vs. B in an indirect comparison as effect A vs. Placebo minus effect B vs. Placebo. Here, we report both direct and indirect interactions of each intervention (see Supplemental Tables 1-6) and the combined effect was calculated and reported. In addition, results of individual studies per treatment were also calculated. The inverse variance was always used in both analyses and results are presented as a standardized mean difference per study, sub-group and network, expressed as Cohen's standardized d (the difference between the means divided by the standard deviation for the data), with 95% confidence interval (95% CI).
In the 30 retained studies, a further search for an oxidative and inflammatory marker was performed. The authors of the retrieved documents use a plethora of parameters to access lipid oxidation and changes in other oxidation and inflammatory parameters. In addition to oxidized LDL (oxLDL) (assessed in 7 studies and analyzed through meta-analysis), all other inflammation and oxidation parameters were reported in a small number of studies. In view of their small number, a meta-analysis was not possible for these markers and therefore a comparison of percent differences was performed with the SPSS V21 program. A significance of 0.05 was retained as a significance threshold.

Description of the Studies
Most of the included studies were randomized cross-over controlled clinical trials with dietary interventions ( Table 1). The characteristics of participants were: healthy subjects in the majority (70%) of the studies. Participants with an established metabolic syndrome were reported in 2 studies and with hypercholesterolemia in 6 studies, while patients with an established cardiovascular or peripheral vascular disease were reported in one study. Postmenopausal women, which were reported in 2 studies, were included in the healthy participant's category. Only males participated in 15 studies and only women in 2 studies, while both sexes were reported in 13 studies. Eleven studies covered a wide range of ages from 18 to 89 years, others (10) focused only on young adults (aged 18-40 years), others (7) examined middle-aged (45-60 years) and the rest of the studies (2) focused on old aged participants >65 years. In most studies (21), the number of participants ranged from 10 to 100 (with exception one study that it was <10), in others (4) the number of participants ranged from 100 to 200 and it was >200 in 4 studies.
The intervention period of most studies (40%), including the washout periods, lasted 2-3 months (30%), 3-12 months (20%), and in 10% the intervention period was over 1 year. The dietary interventions included the consumption of olive oil with a different polyphenol content. In most studies (57%), participants consumed extra-or virgin olive oil high in polyphenols (HPOO), over common or refined olive oil, low in polyphenols (LPOO). In addition, several studies (20%) used extra-or virgin olive oil, in comparison with other oils (corn, fish, etc.).
In most studies (36.6%) participants received a daily dose of 25 ml olive oil and 15 ml in one study. In some studies (16.6%), the participants received a daily olive oil dose of 40 ml (2 studies) and 50 ml (3 studies) and in the rest of the studies (10%), the participants received 68-70 g of olive oil per day.

Effect of Olive Oil on Metabolic Parameters
The network meta-analysis revealed that adherence to the Mediterranean diet significantly decreased circulating glucose levels ( Analyzing further the direct and indirect contribution of the Mediterranean diet adherence and the effect of high and lowpolyphenol content olive oil in these changes (see the Materials and Methods section for details of the used methodology and Supplemental Tables 1-6) it becomes apparent that the main driver for glucose, triglycerides, total and LDL cholesterol is the Mediterranean diet per se, suggesting that this impact on cardiometabolic biochemical indices may be driven by a number of constituents, in addition to those included in the olive oil (see for example the direct effect of HPOO on LDL levels (d = −0.074, as compared to its indirect effect, when the Mediterranean diet is also included (d = −0.116), Supplemental Table 5). In contrast, the effect of olive oil on HDL concentration is directly related to its polyphenol content, as detailed in Supplemental Table 4.    Table 1. HPOO stands for olive oil high in polyphenols, while LPOO presents data of olive oil, low in polyphenols. Standardized differences from a control (westernized) diet are presented, together with their 95% confidence intervals.
In conclusion, a network meta-analysis of olive oil intervention studies in humans suggests that the only direct impact of olive oil polyphenols in cardio-metabolic biochemical indices is its direct effect of HDL-cholesterol levels, while other parameters (blood glucose, triglyceride, LDL) are modified because of the Mediterranean diet, in which olive oil is a major constituent. In addition, our data confirm the beneficial effect of the Mediterranean toward a westernized diet pattern.

Effect of Olive Oil on Oxidative Parameters
An increased resistance of LDL to oxidation is reported in 5/6 studies, a decrease in LDL oxidation rate is found in 3/4 studies, and the detection of antibodies against oxLDL was also decreased by intervention with olive oil ( Table 2). However, no correlation between polyphenol content and the resistance of LDL to oxidation was found. These data suggest that a moderate polyphenol content may be sufficient to protect LDL from oxidation. Further evidence for this is reflected by the decrease of hydroxyl fatty acids and the Lipoprotein lipase gene expression, which increased significantly (+26%) after a high polyphenol olive oil intervention (49), suggesting a direct effect of polyphenols to the transcription of lipid enzyme regulation. These data further verify our network meta-analysis of oxLDL, presented above, which indicates that the major impact on oxLDL is the adherence to the Mediterranean diet, further suggesting that many of the olive oil-attributed benefits might be related to other microconstituents of elements of this diet, in addition to olive oil polyphenols.
DNA oxidation parameters (oxidized DNA bases, DNA breaks and 8-oxo-dG in urine) were decreased by 3-61% after olive oil intervention ( Table 3). Here too, no direct correlation with the polyphenol content was observed, suggesting that a moderate content of polyphenols (calculated by linear regression to 56 mg polyphenols/L olive oil) may be sufficient to prevent nucleic acid oxidation. Corroborating to this effect, the plasma antioxidant capacity (measured by different methods, accessing discrete antioxidant molecules in the plasma/serum) presented an increase after olive oil intervention, indicative of an enhanced role of polyphenols in increasing the organism's antioxidant capacities. Here too, the effect is also observed with low polyphenol content (necessary dose of 98 mg/L), suggesting that other elements may be also interfering with the observed decrease.

Effect of Olive Oil on Inflammatory Markers
High sensitivity CRP decreased in all, but one studies after the olive oil-high polyphenol content intervention ( Table 4). The biggest reductions (−39.4 and −35.86%) in CRP were observed after instauration of a Mediterranean diet profile with the inclusion of olive oil, regardless of its polyphenol content. It  is also to note that in one of these investigations, the target group was patients with a diagnosed metabolic syndrome, while in another one, patients with cardiovascular disease were included. Finally, several cytokines were also decreased after olive oil ± Mediterranean diet intervention.

DISCUSSION
Diets following the principles of the Mediterranean diet, rich in vegetables, legumes, and olive oil have proven their beneficial character in preventing cardiovascular disease (54), diabetes (55)(56)(57)(58), and hyperlipidemia (59). These effects were mainly attributed to the antioxidant fraction of microconstituents present in olive oil and in other elements of this diet (fruits, vegetables). In fact, a meta-analysis of 50 studies and 534,906 individuals revealed that adherence to the Mediterranean diet was associated with reduced risk of metabolic syndrome (60) and their participants expressed lower levels of inflammatory markers related to atherosclerosis (61). The beneficial role of olive oil consumption is nowadays widely recognized, with the European Food Safety Authority (EFSA) approving two health claims regarding olive oil (62). They suggest its use to replace saturated fats to keep normal blood cholesterol levels and protect blood lipids from oxidative stress, with the later effect to be achieved by olive oil polyphenols contained in a daily intake of 20 g of extra-virgin olive oil.
In order to delineate the exact role of olive oil in the above metabolic changes, we have performed a network metaanalysis of 30 human intervention studies. Network metaanalysis evaluates, in addition to the direct effects of each treatment/intervention, the indirect effects resulting from the linear interaction between the network components. For example, the effect of HPOO on glucose levels, shown in the scheme below Supplemental Table 1 is the integrator of its direct effect (line control-HPOO) and the indirect effects (control-LPOO-HPOO and Control-MedDiet-HPOO), while the impact of the MedDiet in the same scheme is its direct effect on glucose (line Control-MedDiet) and the sum of its indirect effects (Control-HPOO-MedDiet and Control-LPOO-MedDiet) (see Salanti et al. (53) and references therein for a thorough description of the Network meta-analysis theory and applications). Using this approach, we show that the effect of olive oil on glucose and circulating lipids cannot be distinguished from the effect of adherence to a Mediterranean diet pattern, while the only clear-cut effect of a high-polyphenol olive oil is its effect on HDL-cholesterol. In this respect, we confirm our recent findings (63) that olive polyphenols, administered in a functional food and at a dose compatible with the EFSAsuggested dose of olive oil, do not modify the circulating glucose levels, while they ameliorate insulin sensitivity. These results were further corroborated by the reported direct protective effect of polyphenols in the pancreas (64) and the amelioration of   insulin secretion through an anti-inflammatory action of oleic acid (65). The effect of olive oil on circulating lipids has been extensively analyzed, in the context of the Mediterranean diet, taking into consideration their impact on cardiovascular diseases (54, 59) and the direct effect of olive oil is clearly demonstrated in performed meta-analyses (14,22). In addition, George et al. (21) reported a moderate reduction of CVD risk by high-polyphenol olive oil, including effects on different oxidative parameters, total, HDL-and oxLDL-cholesterol. However, the authors reported the risk of biases, while the effect of the Mediterranean diet was not accessed. In the PREDIMED (13) and in a recent meta-analysis (13), the effect of HPOO as a protective agent in stroke but not in CVD was reported. Here, we show that the main effect of HPOO is the increase of circulating HDL, while other effects on cardio-metabolic parameters should be attributed to the Mediterranean diet per se. Additionally, the effect of olive oil on triglycerides might be mainly attributed to the lipid fraction than to microconstituents, as it is the same in high and low polyphenol content. These observed effects may be related to the reported impact of oleic acid and its metabolites, as well as of olive oil polyphenols on different enzymes, signaling molecules and a direct effect on the transcription of different proteins, including lipid-related, -transporting, or -metabolizing enzymes (31,(66)(67)(68)(69)(70)(71)(72)(73).
Analyzing further the retained studies, we report (in a narrative review rather than through a meta-analysis, in view of the small number of studies) the effect of olive oil, high or not in antioxidant polyphenols, on oxidation and inflammation parameters. This effect of olive oil polyphenols has retained an increased attention, in view of the deleterious effect of oxidized lipids and nucleic acid damage, related to chronic diseases, including cardiovascular diseases and cancer (31,67,69,71,72,74,75). Analysis of our dataset confirmed the effect of olive oil polyphenols in protecting LDL and nucleic acid oxidation, in accordance to previous meta-analysis (76). However, an unexpected result reported here is that a much lower than the previously reported concentration of olive oil polyphenols is required to induce this protection (∼60 mg/L of olive oil). This finding is in contradiction to the current belief that markers of oxidation (such as oxLDL and nucleic acid oxidation) are inversely related to the polyphenol content of olive oil, while the plasma antioxidant activity is directly related to it. The results of our analysis, if verified in prospective studies, could, therefore, provide a paradigm shift to currently established beliefs and might have a direct impact on the olive oil industry and human nutrition. It is further to note that the effect of olive oil on inflammatory markers is mainly evident in patients with an established metabolic syndrome (27), or in patients with cardiovascular disease (26), providing evidence of a possible protective/therapeutic use of olive oil in such conditions.
The novel element of the present meta-analysis consists of the estimation of LPOO and HPOO effects on cardio-metabolic parameters through a network analysis, estimating both their direct and indirect effects. Our data suggest that the major effect on these parameters is mediated through an adherence to the Mediterranean diet, while the only notable effect of olive oil polyphenols is the increase of HDL-cholesterol and the amelioration of the antioxidant and inflammatory status of the subjects. This effect is more pronounced in subjects with an established metabolic syndrome or other chronic conditions/diseases, evidencing its beneficial health effects. In addition, we report that much lower antioxidant polyphenols may be sufficient for the beneficial effects of olive oil, while we show that the lipid fraction of olive oil may be responsible for some of its beneficial actions. These conclusions, if verified in further prospective trials, may be of value in re-establishing the role of olive oil in human nutrition.

AUTHOR CONTRIBUTIONS
MK and EC conceived and designed the analysis and wrote the paper. ET, NP, and EC performed the analysis. All authors read and approved the final manuscript.