Fruit and vegetable consumption and health outcomes: an umbrella review of observational studies.

Abstract The aim of this study was to provide a comprehensive evaluation of current evidence on fruit and vegetable consumption and health outcomes. A systematic search for quantitative syntheses was performed. Several criteria, including study design, dose–response relationship, heterogeneity and agreement of results over time, and identification of potential confounding factors, were used to assess the level of evidence. The strongest (probable) evidence was found for cardiovascular disease protection; possible evidence for decreased risk of colon cancer, depression and pancreatic diseases was found for fruit intake; and colon and rectal cancer, hip fracture, stroke, depression and pancreatic diseases was found for vegetable intake. Suggestive and rather limited associations with other outcomes have been found. Evidence of potential confounding by sex and geographical localisation has been reported. Despite findings are consistent enough for hypothesising causation (at least for cardiovascular-related outcomes), further studies are needed to clarify the role of potential confounding factors.


Introduction
Dietary recommendations on fruit and vegetable intake are not univocal among countries and organisations, but there is agreement that increasing their consumption may improve general health and decrease the risk of major non-communicable diseases (Slavin and Lloyd 2012). The health-promoting properties of fruit and vegetable depend on several points, including (i) they are rich in vitamins, minerals and antioxidants; (ii) they are generally low in energy density; (iii) they are a valuable source of dietary fibre, (iv) their consumption is often associated with lower intake of unhealthy foods (Liu 2013). Despite the general agreement regarding the role for fruit and vegetable consumption in cancer prevention, evidence is often contrasting and methods of its evaluation highly affect the outcome and the final message to the population. For instance, results from case-control studies are generally consistent toward a general inverse association between fruit and vegetable consumption and occurrence of most of cancers, but data from prospective cohort studies are far more uncertain and the overall beneficial effects toward cancer risk are rather modest (Norat et al. 2014). Also regarding cardio-metabolic outcomes results are contrasting and not often consistent (Woodside et al. 2013). Plant-based diets have been demonstrated to provide potential benefits toward cardiovascular and metabolic health (Kahleova et al. 2017). As fruit and vegetable represents the main food groups characterising such dietary patterns, balanced high quality plant foods may provide adequate intake of micro-nutrients and phytochemicals required to exert such cardiometabolic protective effects (Satija and Hu 2018). Overall, there is scientific and general public demand for more evidence-based review of scientific reports, with comprehensive and systemised approaches to synthetise current evidence and address future directions involving global policies and food systems (Vanamala 2017). Increasing the consumption of fruits and vegetables may be considered an effective strategy for reducing the incidence of chronic diseases, but it is unclear what is the level of evidence of potential protection associated with fruit and vegetable consumption and what are the outcomes more influenced. Thus, the aim of this study was to systematically review the existing evidence from meta-analyses on fruit and vegetable consumption and health outcomes and to assess the level of evidence associated with the findings.

Study selection
A systematic search for quantitative research syntheses of different outcomes investigating the association between fruit and vegetable intake and health outcomes was conducted in Medline and Embase electronic databases up to January 2017. Search terms used were the following: [(fruit OR fruits OR berries OR citrus OR vegetable OR vegetables) AND (meta-analysis OR metaanalysed OR pooled analysis)]. The search was independently performed by two authors (GG and JG) and any discrepancies were solved with discussion. Inclusion criterion was meta-analyses of prospective cohort studies or randomised controlled trials (RCTs) considering fruit or vegetable intake as variable of exposure and any disease condition as outcome. Exclusion criteria were the following: studies exploring the relation between the aforementioned exposures and intermediary biomarkers of disease (i.e. blood lipids, blood pressure, etc.) or intermediary clinical conditions (i.e. variation in body weight/BMI, etc.); and systematic review without quantitative evaluation of the association between exposure and outcome or pooled analysis with selected individual-data studies.

Data extraction
From each meta-analysis included, the following information was abstracted: name of the first author and year of publication, outcome, number of studies included in the meta-analysis, study design of included studies (i.e. case-control/cross-sectional and prospective), total number of population exposed, number of cases, type of exposure, measure of exposure [including highest versus lowest (reference) category of exposure or dose-response incremental servings per day (linear)], effect sizes [risk ratio (RR), odds ratio (OR) or hazard ratio (HR)].

Data evaluation and evidence synthesis
Whenever more than one meta-analysis was conducted on the same outcome, included the same study design, and the same type of population, concordance for the main outcome of interest, including direction and magnitude (overlapping confidence interval) of the association was evaluated. For further analyses, the most recent/exhaustive study was considered. The analyses of the highest versus the lowest (reference) category of exposure and dose-response analyses were evaluated. Direction, magnitude, heterogeneity (I 2 ) and subgroup/stratified analyses for potential confounding factors were considered to have indication of level of evidence. Criteria used for evidence categorisation were modified from the Joint WHO/ FAO Expert Consultation (Cao et al. 2018): briefly, insufficient evidence, when there was availability of meta-analysis including case-control studies, limited prospective cohort studies included in meta-analyses (n < 3), or evident contrasting results from meta-analyses with the same level of evidence; limited/possible evidence, when there was availability of meta-analyses with significant/lack of information on heterogeneity (I 2 > 50%) or identification of potential confounding factors (i.e. different findings in subgroups); probable/ convincing evidence, when there was availability of meta-analyses of prospective cohort studies with no heterogeneity, no potential confounding factors identified and eventual disagreement of results over time reasonably explained/and evidence of dose-response relation further investigated).

Meta-analyses on fruit consumption and health outcomes
The characteristics and summary risk estimates for the highest versus the lowest fruit consumption categories on 20 unique outcomes of 17 non-overlapping meta-analyses including 3 prospective cohort studies (Aune et al. 2011;Aune et al. 2012;Paluszkiewicz et al. 2012;Chen et al. 2013;Hu et al. 2014;Meng et al. 2014;Alsamarrai et al. 2014;Seyedrezazadeh, Moghaddam, et al. 2014;Xu et al. 2015;Fang et al. 2015;Gan et al. 2015;Liu et al. 2016;Luo et al. 2016;Vieira et al. 2016;Wang et al. 2016;Wu, Sun, et al. 2016;Zhan et al. 2017) are presented in Figure 2. The characteristics and summary risk estimates by dose of fruit consumption evaluated in 11 non-overlapping meta-analyses (Aune et al. 2011;Aune et al. 2012;Chen et al. 2013;Hu et al. 2014;Wang, Chen, et al. 2014;Gan et al. 2015;Wu et al. 2015;Liu et al. 2016;Vieira et al. 2016;Wu, Sun, et al. 2016) testing for linear association with 13 unique outcomes are reported in Supplementary Table  1. Meta-analyses on pancreatic diseases, stroke, asthma, lung cancer, depression, cardiovascular disease (CVD), coronary heart disease (CHD), hypertension, colon cancer, colorectal cancer, type 2 diabetes mellitus (T2DM), breast cancer and gastric cancer reported a statistically significant association with reduced risk for the highest versus the lowest category of fruit consumption; those on rectal cancer, hip fracture, bladder cancer, non-Hodgkin lymphoma, prostate cancer and liver cancer showed null associations. Among studies reporting a dose-response analysis, increasing consumption of fruit was linearly associated with decreased risk of stroke, gastric cancer, all-cause mortality, CHD, T2DM, lung cancer and hypertension (Supplementary Table 1). However, among significant findings, studies on stroke, asthma, hypertension and all-cause mortality reported evidence of significant heterogeneity of results between studies. Moreover, some studies showed potential confounding effect by sex (decreased risk of T2DM and colorectal cancer only in women), smoking status (decreased risk of breast and colorectal cancer not significant in studies adjusting for smoking status), BMI, alcohol and education (decreased risk of breast cancer not significant in studies adjusting for these variables) or geographical location (decreased risk of lung cancer, CHD, colorectal cancer and T2DM was not significant in studies conducted in Asian countries; decreased risk of breast cancer was not significant in studies conducted in Asian and European countries; decreased risk of gastric cancer was not significant in studies conducted in Asian countries and United States) (Supplementary Table 2). Decreased risk of lung cancer was significant only among current smokers (Supplementary Table 2).
The list and main findings of meta-analyses for outcomes with more than one meta-analysis published over time are presented in Supplementary Table 3. Comparison of the results by outcome over time showed substantial consistency between results for most of the outcomes. In contrast, results from previous reports on bladder cancer (Yao et al. 2014), colon, rectal and colorectal cancer (Huxley et al. 2009), pancreatic cancer (Alsamarrai et al. 2014), CHD (Law and Morris 1998), and T2DM (Cooper et al. 2012) differed from the latest published. There were no detectable reasons for such inconsistency of results, but the most likely reason was the early years of publication and low number of studies included compared to the latest articles.

Meta-analyses on vegetable consumption and health outcomes
The characteristics and summary risk estimates for the highest versus the lowest vegetable consumption categories on 20 unique outcomes of 59 non-overlapping meta-analyses (Trock et al. 1990;Trock et al. 1990;Law and Morris 1998;Gandini et al. 2000;Steinmaus et al. 2000;Anderson et al. 2000;Smith-Warner et al. 2001;Bosetti et al. 2002;Norat and Riboli 2002;Riboli and Norat 2003;Dauchet et al. 2005;Lunet et al. 2005;Dauchet et al. 2006;Pavia et al. 2006;Bandera et al. 2007;Hamer and Chida 2007;He et al. 2007;Lunet et al. 2007;Huxley et al. 2009;Aune et al. 2011;Aune et al. 2012;Cooper et al. 2012;Paluszkiewicz et al. 2012;Chen et al. 2013;Liu et al. 2013;Johnson et al. 2013;Jin et al. 2014;Li, Jiang, et al. 2014;Yao et al. 2014;Alsamarrai et al. 2014;Hu et al. 2014;Li, Fan, et al. 2014;Li 2014;Liu and Lin 2014;Yang et al. 2014;Wang, Chen, et al. 2014;Ben et al. 2015;Fang et al. 2015;Gan et al. 2015;Meng et al. 2014;Seyedrezazadeh, Moghaddam, et al. 2014;Huang et al. 2015;Liu et al. 2015;Luo et al. 2015;Vieira et al. 2015;Wang, Qin, et al. 2015;Liu et al. 2016;Luo et al. 2016;Wang et al. 2016;Vieira et al. 2016;Wu, Sun, et al. 2016;Wu et al. 2015;Xu et al. 2015;Zhao et al. 2016;Zhan et al. 2017) are presented in Figure 3. The characteristics and summary risk estimates by dose of vegetable consumption evaluated in 10 non-overlapping meta-analyses (Aune et al. 2011;Aune et al. 2012;Chen et al. 2013;Hu et al. 2014;Wang, Chen, et al. 2014;Gan et al. 2015;Liu et al. 2016;Vieira et al. 2016;Wu, Sun, et al. 2016) testing for linear association with 13 unique outcomes are reported in Supplementary Table 5. Meta-analyses on liver cancer, pancreatic diseases, hip fracture, CVD, stroke, CHD, colon cancer, age-related cataract, depression, hypertension, colorectal cancer, and lung cancer reported a statistically significant association with reduced risk for the highest versus the lowest category of fruit consumption; those on pancreatic cancer, non-Hodgkin lymphoma, showed null associations. Among studies reporting a dose-response analysis, increasing consumption of vegetables was linearly associated with decreased risk of colorectal cancer, stroke, all-cause mortality and CVD mortality, CHD, T2DM and lung cancer. However, meta-analyses on liver cancer, hypertension, all-cause and CVD mortality reported evidence of significant heterogeneity of results between studies. Moreover, some studies showed potential confounding effect by sex (hip fracture and T2DM risk was decreased only among men while non-Hodgkin lymphoma was significant only among women) or geographical location (decreased risk of liver cancer and hip fracture was significant only among studies conducted in Asian countries; decreased risk of CHD and hypertension was significant only among studies conducted in the US and Europe; decreased risk of colorectal cancer was significant only among studies conducted in the US; decreased risk of lung cancer was significant only among studies conducted in Europe) (Supplementary Table 6). Decreased risk of lung cancer was significant only among current smokers. No confounding effect by BMI status, alcohol, and education was detected.
The list and main findings of meta-analyses for outcomes with more than one meta-analysis published over time are presented in Supplementary Table 7. Comparison of the results by outcome over time showed substantial consistency between results for hypertension , colon, rectal and colorectal cancer (Huxley et al. 2009), stroke (Dauchet et al. 2005) and CHD (Law and Morris 1998). Possible explanation is the increase in number of studies included in more recent meta-analyses.

Summary of evidence
A summary of the variables (including presence of heterogeneity, dose-response analysis, potential confounding factors and agreement of results over time) investigated to address the strength of evidence from meta-analyses on fruit and vegetable consumption and health outcomes is showed in Supplementary Table 9. The summary evidence from meta-analyses of observational studies on fruit and vegetable consumption and various outcomes is showed in Table 1. Fruit consumption was associated with probable decreased risk of CVD and possible decreased risk of colon cancer, depression and pancreatic diseases; in contrast, evidence on decreased risk of asthma, breast, colorectal, lung and stomach cancer, CHD, hypertension, allcause and CVD mortality, stroke and T2DM was rather limited. There was suggestive but substantially insufficient evidence for the association between fruit consumption and lower odds of colorectal adenoma, upper aero-digestive tract cancer, inflammatory bowel disorders, wheezing and allergies. No substantial evidence of association was found for fruit consumption and the remaining outcomes examined. Regarding vegetable consumption, the strongest evidence (probable) interested decreased risk of age-related cataract and CVD, while there was a possible evidence of decreased risk of colon and rectal cancer, hip fracture, stroke, depression and pancreatic diseases, and limited of liver, colorectal and lung cancer, CHD, hypertension, all-cause and CVD mortality. Insufficient evidence on the association with decreased odds of glioma, endometrial and oesophageal cancer, ulcerative colitis, and wheezing was found and null results for asthma, allergies, colorectal adenoma, various cancers (bladder, breast, naso-oropharynx, NHL, pancreas, prostate, stomach, thyroid), Chron's disease, cancer mortality and T2DM.

Discussion
In this umbrella review, we collected and analysed evidence from existing meta-analyses on fruit and vegetable consumption and various health outcomes. After assessing the strength, the direction and the consistence of the associations, we concluded that the strongest evidence related to fruit and vegetable consumption is a probable reduced risk of CVD and age-related cataract (for vegetable only). Among other possible associations, higher consumption of fruit and vegetable was related with lower risk of colon cancer, depression and pancreatic diseases, while vegetable only was possibly further associated with decreased risk of rectal cancer, stroke (total) and hip fracture. However, the latter associations lacked information on potential confounding factors, making evidence slightly weaker than for CVD.
The level of evidence on the potential protective effect of fruit and vegetable consumption on CVD risk found in our comprehensive revision of literature is in line with the AHA/ACC Guideline to Reduce Cardiovascular Risk and ESC Guidelines CVD prevention (Eckel et al. 2014;Authors/Task Force et al. 2016). Curiously, evidence on cardiovascular-related outcomes other than CVD (i.e. hypertension, stroke, and CVD mortality) had lower level of evidence; the AHA/ASA Guidelines Stroke Prevention (Meschia et al. 2014) and the ACC/AHA Guideline High Blood Pressure (Whelton et al. 2018) suggest that the potential benefits of fruit and vegetable consumption may depend on their content in potassium, which has been demonstrated to be an important factors for endothelial function by increasing potassium re-absorption by kidney and triggering sodium urinary depletion (Adrogue and Madias 2007). The mechanisms underneath the protective role of fruit and vegetables in decreasing the risk of CVD have been widely investigated by considering both the nutrient and non-nutrient components, and their synergy. Vitamins C, E, and b-carotene (vitamin A precursor) and bioactive compounds, like polyphenols, glucosinolates and carotenoids, play an important role in the regulation of the redox status of cells. Dietary micronutrients and mainly polyphenols have been found to decrease risk of mortality due to CVD and to improve endothelial function and vessel protection by inhibition of LDL oxidation from reactive oxygen and nitrogen species (Alissa and Ferns 2017;Grosso et al. 2017a). Among the phytochemicals, polyphenols have been shown to increase nitric oxide (NO) bioavailability, mainly through the inhibition of protein kinase C-dependent NADPH oxidase and the inhibition of the vasoconstrictor endothelin-1, which in turn lead to blood pressure reduction and inhibition of platelet aggregation, secretion and adhesion, which are at the basis of the causes of strokes or thrombosis (Pignatelli et al. 2006;Flammer et al. 2013). Furthermore, it has been showed that a diet rich in fruit and vegetables (endorsed as 'high total antioxidant capacity' diet) significantly correlates with a decrease of biomarkers of low-grade inflammation (i.e. high sensitivity-C reactive protein, hs-CRP), which has been associated with increased risk of CVD (Brighenti et al. 2005). It has been hypothesised a synergic role of polyphenols and vitamins in affecting the subclinical inflammatory status through the modulation of the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-KB) DNA binding activity. In fact, NF-KB activation seems to be primarily induced by the increased level of oxidative stress, counteracted by phytochemicals and vitamins of fruit and vegetables, leading to increased production of cytokines and chemokines in the vascular endothelium, which in turns stimulate the liver hs-CRP synthesis (Brighenti et al. 2005). Another potential mechanism implicated in the protection from the CVD risk may involve the role of the dietary fibre (Lunn and Buttriss 2007). For example, soluble fibre let the formation of a denser gelatinous chyme-due to its property to attract liquids from the interstices-leading to delay the gastric emptying, maintain higher level of satiety and decrease the amount of ingested foods. Once at the small intestine level, the produced gel treads slowly the duodenal tract, decreasing the time of nutrients absorption and leading to a slower glucose absorption, which in turns smooths the post-prandial glycemic peak, one of the main risk of diabetes insurgence. Finally, the chyle and the insoluble fibre reach the colic tract and the latter is substrate of gut microbiota strains-mainly Bifidobacteria-which degrade the long indigested polysaccharide chains and release in the bloodstream short chain fatty acids (SCFA). These short fatty acids-in particular propionic and butyric acid-have been found to inhibit the b-Hydroxy b-methylglutaryl-CoA reductase activity, with the consequent hint of the cholesterol synthesis (Wong et al. 2006). The main putative mechanisms linking the consumption of fruits and vegetables to the decreased risk of CRC are mainly related to the modulation of the apoptotic processes and inhibition of the inflammatory cascades at epithelium level. In vitro studies revealed that flavonoids, i.e. quercetin and kaempferol, both contained in broccoli, apples, onions, grapes, etc., inhibit the b-catenin/T-cell factor (Tcf)/lymphoid enhancer factor (Lef) signalling, which cause the transcription of downstream genes such as cyclin D1, myc, vascular endothelial growth factor (VEGF) and matrix metalloproteinases (MMPs), involved in tumorigenesis along the gastrointestinal tract (Park and Choi 2010); quercetin and kaempferol have been shown to decrease gene expression of cyclin D1 and to induce apoptosis through the activation of key factors, like p21 and p53 (Pan et al. 2011). Similar mechanism has been also attributed to proanthocyanidins-mainly present in dark fruits-through the inhibition of the altered functionality of Raf/MEK/mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3 kinase (PI3K)/Akt pathways, leading to apoptosis processes of cancerous cells (Engelbrecht et al. 2007). Carotenoids, lutein and astaxanthin, mainly contained in red/orange fruits and vegetables, have been found to block the colon cell carcinogenesis processes by modulating the inflammatory processes led by the uncontrolled expression of cyclooxygenase (COX)-2 and MMPs (Nagendraprabhu and Sudhandiran 2011) as well as by inhibiting NF-KB, which in turn promotes the activation of the caspase-3 cascade and trigger the apoptosis processes (Palozza et al. 2003). Dietary fibre has also been endorsed as modulator of the carcinogenesis processes due to (i) fermentative products, which have been found to have a role in the inhibition of TNF-a production, NF-jB activation, and IL-8, -10, -12 expression, and (ii) modulation of the gut microbiota, i.e. by stimulating the growth of beneficial Bifidobacteria strains and limiting the selection of pathogenic strains, such as E.coli, Salmonella and Listeria (Zeng 2014). Besides, dietary fibre increase the viscosity and faecal bulking, allowing the potential dangerous fermentative products to be in contact with the colon-rectal epithelium for a shorter time (Zeng 2014).
In normal colonocytes, butyrate is utilised as the primary energy source to maintain homeostasis. As it is readily metabolised in the mitochondria via fatty acid oxidation, relatively little accumulates inside the nucleus. On the other hand, in cancerous colonocytes, glucose is the primary energy source due to the Warburg effect. Butyrate is still transported into the cell via monocarboxylate transporters but is not metabolised in the mitochondria, which allows it to accumulate in the nucleus and function as an HDAC inhibitor to epigenetically regulate genes involved in cell proliferation and apoptosis (Bultman 2017).
Several different mechanisms have been proposed to explain the inverse association between fruit and vegetables intake and depression development. One of this endorses the antioxidant and anti-inflammatory content of fruit and vegetable (i.e. vitamins and polyphenols) for its role of radical scavengers against the inflammatory processes that have been associated with occurrence of depression (Payne et al. 2012). The intake of vitamin C, carotenoids and polyphenols from fruit and vegetables have been found to counteract the radical species created by intense aerobic activity of brain as well as to hint the oxidation of neuronal membrane phospholipids due to radical species (Payne et al. 2012). Depression has been also defined as a metabolic disorder and associated to obesity, metabolic syndrome and type-2 diabetes (Lang and Borgwardt 2013). The impact of fruit and vegetable fibre on the gastric emptying and satiety sense directly affects leptin, neuropeptide Y, gastric inhibitory peptide (GIP) and brain-derived neurotrophic factor (BDNF), key signals in the development of anxiety, depression and, consequently, weight gain and obesity-related diseases (Lang and Borgwardt 2013). The inflammation, which is triggered during metabolic diseases, seems to be boosted in depressive diseases: diets with high glycaemic loads and poor in fruit and vegetables have been found to increase the hs-CRP levels as well as circulating cytokines and interleukins plasma levels (Milaneschi et al. 2012). Folates, also called B9 vitamin group, are a group of vitamins present as in the oxidised (folic acid) as in the reduced (5-methyltetrahydrofolate, 5-MTHF) form in vegetables, i.e. lettuce, broccoli, asparagus, spinach. They are responsible for the donation of methyl groups for the synthesis of nucleosides, some amino acids and, mainly, for the conversion of homocysteine in methionine and cysteine (with the co-presence of vitamin B12). This last aspect as been associated to depression risk: higher levels of homocysteine and low levels of 5-MTHF have been associated to increased cerebrovascular inflammation, impairments in brain methylation reactions and affection in concentrations of monoaminergic neurotransmitters, i.e. serotonin, dopamine and noradrenaline (Ara ujo et al. 2015). Intriguingly, it has been proved that a diet lacking in folates has a key role in the missing response to drug administration for the treatment of depression, meaning the importance of adequate circulating levels of dietary folates (Papakostas et al. 2005).
Pancreatic disorders might affect the exocrine physiology, with pancreatitis distressing the digestive and the inflammatory processes, or the endocrine part, leading to glycaemia-related disturbs. Concerning pancreatitis, the triggering agent seems to be an uncontrolled auto-digestion of the pancreas cell by trypsin and the consequent altered fibrotic tissue alters the pancreas physiology. Almost all the research works agree in finding the upcoming unbalanced redox state of the pancreatic cell as well as the acute inflammation as the main source of radical species. Elevated level of superoxide dismutase and low concentration of vitamin C have been found in patients suffering of pancreatitis (Waldthaler et al. 2010;Oskarsson et al. 2013). For these reasons, the consumption of fruit and vegetables lead to the ingestion of dietary antioxidants, (e.g. n-acetylcysteine, polyphenols, vitamin C, vitamin E and selenium) able to counteract the elevated amount of oxygen and nitrogen radical species (Siriwardena et al. 2007).
The results on risk of fractures are in line with the report of the National Osteoporosis Foundation (Weaver et al. 2016). Oxidative stress enhances the osteoclastogenesis, inducing apoptosis of osteoblasts and osteocytes, and causing bone resorption and increasing the risk of fracture (Cao et al. 2018). b-carotene and lycopene have been found to suppress osteoclastic differentiation, promote osteoblast metabolism, and stimulate the alkaline phosphatase activity of osteoblasts (Rao et al. 2007;Cao et al. 2018). These mechanisms are mainly controlled by the ratio between the anti-osteoclastogenic osteoprotegerin (OPG) and the pro-osteoclastogenic receptor activator of NF-KB (RANK) ligand. Among phytochemicals, not only isoflavones, but also quercetin and kaempferol play a role in influencing this ratio towards the bone mineralisation (Trzeciakiewicz et al. 2009). Concerning minerals, fruit and vegetables are sources of calcium, potassium and magnesium, which have been shown to positively affect the OPG/RANK ratio and renal mineral reabsorption (Blaine et al. 2015). Vitamin K, which is abundantly present in green leafy vegetables, has a key role in the c-carboxylation of osteocalcin, a protein synthesised by osteoblast which is implicated in bone mineralisation and calcium ion homeostasis (Tucker 2009). Finally, the debated acidbase theory hypothesises that fruit and vegetables are rich sources of alkaline precursors, like calcium, potassium and magnesium and their intake, together with the consequent bone deposition, is useful to counteract the calciuric effects of acids derived from the diet (Qiu et al. 2017).
Regarding the other outcomes, most of retrieved evidence is of possible decreased risk, but heterogeneity between studies and potential confounding factors cannot be ignored and should be further investigated with in-depth studies. Among other possible associations between fruit and vegetable consumption and human health, evidence of potential protection against some cancers (mainly colon) has been found. Healthy dietary patterns rich in fruit and vegetables have been consistently associated with decreased risk of certain cancers (Grosso et al. 2017b;Schwingshackl et al. 2017). There is epidemiological evidence that bioactive compounds, such as polyphenols, may play a role in cancer prevention (Grosso et al. 2017c). Together with polyphenols, also carotenoids and glucosinolates can induce the activation of detoxification phase I and II systems (mainly in liver and gut), leading to a rearrangement and inactivation of the instable carcinogens either by blocking the formation of adducts with DNA or by counteracting their oxidative role (De Kok et al. 2008). Detoxification systems, i.e. glutathione S-transferase (GST), NAD(P)H:quinone reductase and c-glutamylcysteine synthetase, are able to catalyse the formation of less-reactive water-soluble conjugates from carcinogenic compounds (Johnson 2007). Brassicaceae are a vegetable family rich in bioactives, among which glucosinolates. These compounds are biologically inactive but, after chewing and ingestion, myrosinase enzyme converts these compounds in their biologically active form, called isothiocyanates (Angelino and Jeffery 2014). which have been showed to enhance the activity of the detoxification enzymes and to induce their transcription from nuclear DNA. In fact, the isothiocyantes of thiol groups are able to destabilise the ubiquitination of the Kelch-like ECH-associated protein-1 (KEAP-1) and the turnover of the transcription factor nuclear factor, erythroid 2 p45-related factor 2 (Nrf2). The dimer Nrf2-KEAP-1 is able, together with maf proteins, to vehicular into the nucleus and bind the Antioxidant Responsive Element (ARE) genes. This leads to an enhancement of the activation of several different genes responsible for the transcription of several detoxification enzymes or intracellular antioxidants, such as glutathione or thioredoxin (Talalay et al. 1995;Dinkova-Kostova and Abramov 2015). These mechanisms have been found in several different tissues, among which lung, breast and stomach (Dinkova-Kostova et al. 2017). Bioactive compounds from fruit and vegetables have a key role also in triggering the suppression of the tumour progression mainly by interfering with the intracellular signals factors. One of the main important pathway to be influenced is NF-kB, which is able to regulate the expression of hundreds of genes and the consequent transcription of protein deeply involved in inflammation cascades, as above mentioned. Among these, the B-cell lymphoma 2 (BCL-2) proteins are a family of protein which can exert a pro-or anti-apoptotic role on the basis of the conditions. In several cancer cell models, including lungs and breast, bioactive compounds like polyphenols and glucosinolates are able to play as pro-apoptotic compounds and initiated a double way mechanism: (i) to inhibit anti-apoptotic proteins, like BAD and BLK; and (ii) to activate proapoptotic proteins, like BAX and BAK (Surh 2003;Ramos 2008). The result of these actions is the enhanced release of cytochrome C from the mitochondria to cytosol, and the consequent activation of the caspase family enzyme. Particularly, cytochrome c forms an oligomeric complex of cytochrome c/Apaf-1/caspase-9, the so called "apoptosome", which activates the initiator caspase-9 to subsequently cleave the effector caspases-3 and -7, destabilise the DNA chain, and induce cell suicide through the p53-dependent pathway (Scorrano and Korsmeyer 2003;Forbes-Hernandez et al. 2014). Despite mechanisms at cellular level have been provided, evidence for certain cancer, such as lung, breast, and stomach, is still limited mainly due to heterogeneity between studies or potential confounding factors (Grosso et al. 2017d). The concerns regarding the weak evidence is in line with the conclusions of the World Cancer Research Fund (WCRF) Continuous Update Project (CUP), underlining lack of linear association and high heterogeneity (especially in relation with smoking status) (2018. 2018a, 2018. 2018b). Further studies are needed to improve evidence on cancer risk.
Regarding the potential effects of fruit on T2DM, evidence was null for vegetables and rather limited for fruit; several associations, including the IDF and the American Diabetes Association, did not report strong statements on the potential protective effects of fruit consumption toward T2DM risk, rather referring to an overall "healthy diet" including plant-based foods (American Diabetes 2018; International Diabetes Federation 2012). Most of the mechanisms focus on the putative regulation of insulin secretion and sensitivity by dietary fibres and antioxidants. Among the formers, mainly soluble fibre enhances the formation of the viscous gel-forming chyle, which in turns hinders or delays the carbohydrate absorption, resulting in lower postprandial glucose and insulin spikes (Weickert and Pfeiffer 2018). Besides, based on the evidence that a low glycaemic index (GI) diet significantly decrease the insulin resistance and post-prandial blood glucose concentrations, this mechanism is enhanced when fruits with low GI are consumed. e.g. apples, oranges, pears or berries, in spite of high GI ones, e.g. tropical fruits, bananas or grapes (Du et al. 2017). The dietary fibre intake-and the increasing of chyle viscosity-also indirectly affects the incidence of T2D by delaying the gastric emptying, increasing the satiety and decreasing the subsequent hunger. All these effects lead to the decreasing of body weight and waist circumference, one of the main risk factor for insulin resistance and T2D. The second indirect effect of soluble dietary fibre on T2D is related to their gut microbiota fermentation once in the low gastrointestinal tract. In fact, in vitro and in vivo studies pointed out that SCFA derived from dietary fibre-mainly propionic and butyric acids-play a key role in the inhibition of the cholesterol synthesis, the promotion of adipogenesis and the production of leptin, other than the modulation the hepatic glucose metabolism (Canfora et al. 2015). Dietary fibre is also carrier of antioxidants that have been found to negatively affect the T2DM risk. Among the main classes, anthocyanins (contained in red to violet fruits), flavan-3-ols (contained in berries, grapes, etc.), and flavanones (contained in citrus fruits) showed, in in vitro and in vivo studies, strong a-glucosidase and a-amylase inhibitory activities at the intestinal brush border level, leading to a delay/inhibition of the carbohydrate digestion. More, in intestinal cell culture models, they have been found to inhibit glucose transporters across the membrane, among which sodium-dependent glucose transporter (SGLT)-1 and GLUT2. Finally, it has been demonstrated that such polyphenols are able to stimulate the production of insulin induced by glucose, indicating a pivotal role in the amelioration of the insulin sensitivity (Scalbert et al. 2005;Hanhineva et al. 2010).
Limited evidence from case-control studies has been also found for high consumption of fruit and vegetable and decreased odds of inflammatory bowel diseases (such as Chron's disease and ulcerative colitis) and wheezing/asthma/allergies, despite the risk of the latter was significantly lower only in association with fruit consumption. Inflammatory bowel diseases, such as ulcerative colitis and Crohn's disease, are characterised by a common chronic inflammationwith different symptoms, time-laps and anatomical tracts-and a dysbiosis of the gut microbiota due to the inflammatory status, the decrease of the bacterial diversity and the increasing of non-beneficial bacteria strains. The role of fibres and antioxidants has been associated as gut microbiota inflammatory cascades modulators. In fact, the microbial fermentation of soluble fibre-particularly present in fruit and vegetables instead of cereals-produce SCFA, among which butyrate is known for being an important energy source of colonocytes and inhibit inflammation and enhance the barrier function (Rajilic-Stojanovic et al. 2015). Studies on cavies revealed that soluble fibresand other bioactives, such as indol-3-carbinol from cruciferous vegetables-are agonists of the aryl hydrocarbon receptor (AhR), widely expressed in intestinal intraepithelial lymphocytes, responsible for the activation of the IL-22, a potent mediator of cellular inflammatory responses (Ananthakrishnan et al. 2013). Polyphenols have been widely studied for their beneficial effects in prevention and amelioration of inflammatory bowel diseases symptoms, as they are able to counteract both the inflammation and the oxidative stress cascades. Polyphenols are able to inhibit the expression of TLR4 receptor, the activation of IKK and MAP kinases as well as the activation of PPAR-c, resulting in the inhibition of NF-KB activation. Together with the polyphenol modulation of Nrf2, these events lead to the inhibition of inflammatory signals (i.e. chemoattractant proteins, cyclooxygenase-2, iNOS, etc.) and the enhancement of the scavenger activity against radicals by neutralising the single radicals or by activating the endogenous antioxidant enzyme systems (Farzaei et al. 2015).
Mechanisms potential influencing asthma, allergies and wheezing related to fruit intake have been related to the antioxidant and immunomodulatory activities of bioactive compounds, such as vitamins (including vitamin A and C) and polyphenols (including flavanones, flavan-3-ols and anthocyanins) that seems to have a scavenger activity towards the high levels of free radicals produced in the frame of the inflammation reactions. Besides, they are able to enhance the endogenous antioxidant systems, such as glutathione-S-transferase enzyme (Seyedrezazadeh et al. 2014). Among other compounds, vitamin E, quercetin and selenium, are able to inhibit the cytokines production by the T-helper-2 cells. Particularly, those bioactive compounds inhibit the bind of NF-KB to the activator protein 1 of the promoter region of IL-4 gene, inhibiting IL-4 mRNA transcription, which is a trigger factor for other white cells and inflammatory cascades (Devereux and Seaton 2005).
The present study has some limitations that should be addressed. First, the results shown in this report share the common issues of the original meta-analyses included through the systematic search, such as: (i) lack of homogeneity in measurement methods (i.e. food frequency questionnaires vs. dietary recalls), (ii) disagreement in quantification of serving among studies, (iii) lack of information regarding type of vegetable (i.e. leafy green, yellow, starchy, etc.), method of cooking (raw, boiled, fried canned, pickled, etc.) and method of consumption (i.e. fruit juices, smoothies, soups, casseroles, stews, etc.). Second, limitations regarding this specific umbrella review are due to lack of data from long-term RCTs aimed to assess a causal relation between fruit and vegetable exposure and the investigated outcomes; in fact, existing RCTs investigate intermediate biomarkers of disease and were not taken into account for this review by study design. Third, fruit and vegetable consumption is generally a health-conscious choice, which tend to cluster with lower prevalence of smoking and higher physical activity habit (Grosso et al. 2017c): thus, uncontrolled or residual confounding cannot be excluded definitively and might explain the different associations observed for certain outcomes in relation to the sex and country of living of participants. Fourth, relevant quantitative data on publication bias evaluation (i.e. Egger's test) has been searched in the included metaanalyses; however, most of meta-analyses provided p-values only for pooled studies including fruit and vegetable exposure or studies with different design (i.e. prospective and case-control studies). Due to the limited availability of such data, information on publication has not been included in this study and we cannot rule out the possibility of its influence in defining the overall evidence of association between fruit and vegetable consumption and health outcomes. Finally, the evaluation of study quality was not considered in the design of this umbrella review.
In conclusions, fruit and vegetable consumption has been shown to provide substantial benefits toward human health. Despite the findings are quite consistent and there is evidence for hypothesising causation, at least for CVD, further studies are needed to clarify the potential confounding effect of sex and, particularly, geographical localisation.