A large body of epidemiological evidence strongly suggests a primary role for plant-based dietary patterns in reducing the risk of diseases(Reference Serafini, Jakszyn and Luján-Barroso1). However, the identification of the molecules involved in the protective effect of vegetable food and their mechanism of action is far from being understood(Reference Key2). Phytochemicals contained in the plant kingdom are hypothesised to reduce free radical-related cellular damage, potentiating redox defence of the body and contributing to reduction of the risk of developing oxidative stress-related diseases(Reference Khan, Afaq and Mukhtar3). On the basis of the so-called ‘antioxidant hypothesis’, long-term clinical trials have been developed to investigate the effect of supplementation with natural antioxidants to reduce the development of oxidative stress-related diseases. However, clinical trials have produced extremely contrasting results, highlighting the difficulties in mimicking the healthy effect of plant-derived foods through the usage of natural supplements(Reference Bjelakovic, Nikolova and Gluud4, Reference Bjelakovic, Gluud and Nikolova5). Moreover, a negative effect in increasing mortality rates of overall mortality and cancer has been observed in specific meta-analyses, raising strong concerns about the use of galenic antioxidants for disease prevention(Reference Bjelakovic, Gluud and Nikolova5, Reference Bjelakovic, Nikolova and Gluud6).
Antioxidant defences of the body are composed of molecular and enzymatic players; however, the composition of this network markedly differs, in terms of concentration and components, between body compartments(Reference Serafini, Villano and Spera7). The multifunctional properties of the antioxidant network highlight the crucial importance of dynamic interactions among the components of the network in protecting body fluids from oxidative stress. Non-enzymatic antioxidant capacity (NEAC), defined as the moles of oxidants neutralised by 1 litre of plasma, represents a biomarker measuring the antioxidant potential, including synergistic interactions, of body fluids and tissues(Reference Serafini, Villano and Spera7). In terms of participation of individual components to the network, the main contributor to NEAC is uric acid (40–55 %), followed by thiol groups (10–24 %), ascorbic acid (8–15 %) and vitamin E (less than 10 %)(Reference Serafini, Miglio and Peluso8). The contribution of single antioxidants to overall NEAC leaves unexplained a percentage ranging from 20 to 40 % that might be accounted for by dietary components and/or synergistic interactions. NEAC methodology has been widely used to investigate the role of diet in modulating antioxidant function in humans, though criticisms have been raised on the real ability of NEAC to be a proper biomarker of antioxidant function in vivo (Reference Serafini, Miglio and Peluso8, Reference Bartosz9). Moreover, the large number of assays available for the measurement of NEAC in biological matrices has become a major obstacle in understanding the pros and cons of NEAC as a biomarker for assessing antioxidant capacity in vivo.
The main point of concern relies on the lack of a clear association between the ingestion of plant foods rich in antioxidants and changes in plasma NEAC. Plasma NEAC was shown to respond to the ingestion of dietary food items rich in antioxidants such as tea, wine, chocolate, fruit and vegetables in the majority of acute ingestion studies(Reference Serafini, Miglio and Peluso8). However, if we analyse chronic intervention studies, the picture is much more complex and results are not homogeneous. Recently, it has been suggested that the efficiency of antioxidants in modulating NEAC in chronic trials depends on the ‘healthy status’ of the subjects, with larger effects occurring in subjects characterised by risk factors (i.e. smoking), pollutant exposure(Reference Barbato, Tomei and Tomei10) or pathologies (HIV+ and CVD)(Reference Serafini, Miglio and Peluso8). Healthy subjects might have a lower responsiveness to antioxidant supplementation because they do not have any oxidative stress conditions in which a higher intake of redox molecules might be required.
In order to clarify these concerns, we performed a meta-analysis of human intervention studies testing the effect of vegetable food and beverage consumption on plasma NEAC. Moreover, the effect of oxidative stress-related risk factors (OSRRF) and of the applied methodology was also investigated.
Experimental methods
Search strategy
An extensive search for ‘antioxidant capacity’, ‘antioxidant potential’, ‘antioxidant status’ and diet or specific food/beverages (fruit, fruit juice, vegetable, tea, chocolate and wine) was performed, collecting dietary intervention studies up to May 2010. Papers suitable for inclusion were identified by systematic research on the MEDLINE and EMBASE electronic databases. Searches were limited to human subjects and results reported in the English language. In addition, a manual search of references from reports of clinical trials or review articles was performed to identify relevant studies.
Study selection
In the present meta-analysis, we included dietary intervention studies with an experimental design (EXP) as well as with a quasi-experimental design (Q-EXP). EXP studies were randomised, with a control group and a parallel or cross-over design. Observational studies (pre- and post-intervention or pre- and post-data), non-randomised or uncontrolled study design were included in the Q-EXP study category(Reference Harris, Bradham and Baumgarten11). Q-EXP studies were pooled together with EXP studies only after assessing whether they were in agreement with EXP studies(Reference Harris, McGregor and Perencevich12, Reference Borenstein, Hedges and Higgins13).
General inclusion criteria were as follows: (1) human intervention studies assessing the effectiveness of foods, beverages or dietary patterns on plasma NEAC; (2) studies with both acute and chronic supplementation periods; (3) chronic studies with a treatment period longer than 7 d; (4) healthy people and subjects characterised by OSRRF (smoking, dyslipidaemia, heart transplant, obesity and ageing) or pathologies (HIV+ and CVD); (5) numerical or graphical information about results, study duration (chronic studies) or time point (acute studies) and dosage of the plant-derived foods and beverages used. Studies were excluded when only abstracts were available; experimental meal did not fall into the categories that we considered; data presentation was incomplete; treatment effects on NEAC were reported as AUC; subjects were < 18 years; information about the dose of food and beverages was incomplete; the washout time was not considered or was less than 7 d (only for the cross-over design).
When necessary, efforts were made to contact investigators for clarification or additional data.
Quality assessment and data extraction
For data extraction, two investigators (D. L.-B. and G. M.) independently screened the titles and abstracts resulting from the search strategies. Articles with titles or abstracts clearly irrelevant were rejected at first screening. The full text of potentially relevant articles was reviewed to assess eligibility for inclusion in the meta-analysis with any disagreement resolved by consensus. Considering that the present meta-analysis included different types of studies, it was performed following both PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) and MOOSE (Meta-Analysis of Observational Studies in Epidemiology) checklists(Reference Manchikanti, Datta and Smith14).
Quality assessment of the included studies was performed using study design as the main quality variable (parallel, cross-over studies or pre–post-intervention studies, placebo, studies with or without a control group). Moreover, other features related to human dietary intervention studies, such as the number of participants enrolled, the length of studies for chronic ones and the description of experimental meal, were considered. Score quality was not used as a criterion for the selection of trials, but it was only used for allocation purpose in high-, fair or low-bias risk. Trials with an adequate study design were classified as low-bias risk (good quality: >7 points), whereas those with an adequate design but with one or more unclear quality components were classified as fair quality (4–7 points). Trials with an inadequate study design were considered as high-bias risk or poor quality ( < 4 points; Tables 1 and 2). Extracted data include: name of the first author; year of publication; experimental meal; number of participants; health status of the enrolled subjects; type and design of the study and time point considered for data analysis; daily dosage; assay used; study quality.
Table 1 Criteria utilised to assess the quality of the studies included in the meta-analysis
Table 2 Quality score used in the meta-analysis
Data analysis
Relevant data of the studies available for formal meta-analytic evaluation were entered into Comprehensive Meta-Analysis software (Biostat) for evaluation of the meta-analysis.
In particular, for parallel or cross-over studies, when results were reported as post-data only, we used mean, standard deviation and sample size in each group, or difference in means, sample size and P value between groups. When results were reported as pre- and post-data, we used mean, standard deviation, sample size in each group and correlation between baseline and end-point intervention period, or mean change, standard deviation difference, sample size in each group, correlation between baseline and end-point intervention period. For observational studies considering only one group (pre–post-intervention data), we used mean difference, standard deviation of difference and sample size. For all data, standard deviations were calculated, when necessary, from standard error or CI, and data not provided in numerical form were estimated from figures. In all types of dietary intervention studies, we assumed the correlation between baseline and end-point study period to be 0·5 to produce the most conservative estimate(Reference Borenstein, Hedges and Higgins13, Reference Higgins15).
When in the same study the effect on plasma NEAC of two different doses of the same experimental meal was evaluated, our approach was to create a synthetic variable for each study to compute a combined effect based on these variables (defined as the mean of the effect size), and we used it as the effect size in the analysis. The same procedure was used in chronic studies that reported the effect on NEAC of the same experimental meal, in two or more periods, as well as in acute studies that showed the effect on NEAC of the same experimental meal, in two or more time points.
For acute studies, the significant change (positive or negative) on plasma or serum NEAC was considered as the time point to statistical analysis, while if there were no reported significant changes, we have considered the time point at the end of the study. For chronic studies, we considered the end point of the intervention.
All studies addressed the same outcome (plasma antioxidant capacity or NEAC), used different methods of analysis (total radical-trapping antioxidant parameter (TRAP), ferric-reducing antioxidant power (FRAP), oxygen radical absorbance capacity (ORAC) and Trolox equivalent antioxidant capacity (TEAC)) and presented the results in different scales. To enable a joint comparison, the same index of treatment effects was used, and in each individual study, the standardised mean difference (SMD) with 95 % CI was calculated(16). In the present analysis, a positive SMD indicates an increase in plasma NEAC after consumption of plant-derived foods and beverages (experimental meal). When an author published the plasma NEAC value detected with two or more methods, we evaluated separately the effect size in relation to the assay used.
Meta-analysis was performed using DerSimonian & Laird's ‘random-effects models’. Under the random-effects model, effect sizes are assumed not to be common to all studies; they vary under a normal distribution model(Reference DerSimonian and Laird17). We used a random-effects model because there was evidence of heterogeneity between studies based on the χ2 test for heterogeneity at the significant level of P= 0·10 and the I 2 statistic, percentage of total variation across studies due to heterogeneity, rather than chance, of more than 50 %(Reference Higgins and Thompson18).
Potential publication bias (i.e. the association of publication probability with the statistical significance of study results) was investigated using visual assessment of the funnel plot calculated by Comprehensive Meta-Analysis software(Reference Sterne, Gavaghan and Egger19) (Biostat).
Results
We identified 716 potential literature citations through the database searches (Fig. 1). At first, 156 studies met our inclusion criteria and five additional studies potentially meeting our inclusion criteria were identified through scanning reference lists. Of the 161 references, seventy-three papers were excluded for the following reasons: outcome measures not suitable for meta-analysis, continuous variables of pre- and post-intervention data or mean difference for only post-data not included in the publication; NEAC values expressed as AUC; chronic intervention trials were less than 7 d or intervention studies performed with dietary supplements. The remaining eighty-eight studies, reporting a total number of 122 interventions met our selection criteria and were included in the meta-analysis (Table S1, available online).
Fig. 1 Flow chart of the study identification and selection.
The following three main categories of intervention, on the basis of type of food, were identified: beverages (n 47); plant foods (n 34); plant food-based dietary patterns (n 7). Considering the criteria used to assess the quality of studies included in the present meta-analysis, thirty-six studies were of good quality, fifteen studies were fair and thirty-seven studies were of poor quality. Moreover, forty of the included studies had an experimental design (EXP), whereas forty-eight studies had a quasi-experimental design (Q-EXP). Plasma NEAC was assessed in 2890 subjects. Moreover, all subjects were stratified into two main subgroups: healthy subjects (n 1880) and subjects with risk factors or diseases (n 1010). After evaluating the agreement between Q-EXP and EXP studies through the view of effect size direction (data not shown), both Q-EXP and EXP studies were processed together(Reference Borenstein, Hedges and Higgins13).
Beverages
Intervention studies involving red wine (n 16), fruit juices (n 20) and tea (n 12) are described in Fig. 2. There was clear evidence on the efficacy of red wine consumption in increasing plasma NEAC, both in acute feeding studies (SMD 1·118, 95 % CI 0·976, 1·325; P< 0·000) and in chronic studies (SMD 1·130, 95 % CI 0·818, 1·617; P< 0·001). Fruit juice ingestion was able to increase plasma NEAC in acute feeding studies (SMD 1·028, 95 % CI 0·446, 1·702; P= 0·001), but was ineffective in chronic intervention studies (SMD 0·157, 95 % CI − 0·275, 0·422; P= 0·382). Tea consumption induced a similar increase in plasma NEAC both after acute (SMD 0·703, 95 % CI 0·154, 1·248; P= 0·012) and chronic ingestion (SMD 0·664, 95 % CI 0·249, 1·078; P= 0·002). The results showed that green tea had a stronger antioxidant effect than black tea (SMD 0·957, 95 % CI 0·447, 1·466; P< 0·0001 for green tea and SMD 0·487, 95 % CI 0·072, 0·901; P= 0·021 for black tea) (data not shown).
Fig. 2 Effect of beverage ingestion on plasma non-enzymatic antioxidant capacity (NEAC). Studies were stratified according to the design of the study (acute or chronic). A positive standardised mean difference (SMD) indicates an increase in plasma NEAC, whereas a negative SMD indicates a decrease in plasma NEAC. The terms a, b or c was used when the same author measured plasma NEAC with different methods. (A colour version of this figure can be found online at http://www.journals.cambridge.org/bjn).
Plant foods
The effect of chocolate (n 11), fruits (n 10) and vegetables (n 12) on plasma NEAC is described in Fig. 3. Chocolate increased plasma NEAC in acute studies (SMD 0·677, 95 % CI 0·509, 1·066; P< 0·008), but failed to enhance plasma NEAC after chronic ingestion (SMD − 0·034, 95 % CI − 0·540, 0·473; P= 0·897). Similarly, antioxidants from fruits massively increase endogenous antioxidant defences in acute studies (SMD 1·260, 95 % CI 0·946, 1·535; P< 0·000), while no effect was revealed in chronic feeding studies (SMD 0·508, 95 % CI 0·133, 0·938; P= 0·105). On the other hand, vegetables induced a significant increase in NEAC both after acute (SMD 0·629, 95 % CI 0·141, 1·117; P< 0·012) and chronic consumption (SMD 0·383, 95 % CI 0·120, 0·647; P= 0·004).
Fig. 3 Effect of plant food ingestion on plasma non-enzymatic antioxidant capacity (NEAC). Studies were stratified according to the design of the study (acute or chronic). A positive standardised mean difference (SMD) indicates an increase in plasma NEAC, whereas a negative SMD indicates a decrease in plasma NEAC. The terms a, b or c was used when the same author measured plasma NEAC with different methods. (A colour version of this figure can be found online at http://www.journals.cambridge.org/bjn).
Plant food-based dietary patterns
In agreement with the results observed for fruit and vegetables, dietary patterns based on fruit and vegetables had a strong impact in modulating positively plasma NEAC (SMD 0·965, 95 % CI 0·809, 1·122; P< 0·000; Fig. 4).
Fig. 4 Effect of chronic ingestion of plant food-based dietary patterns on plasma non-enzymatic antioxidant capacity (NEAC). A positive standardised mean difference (SMD) indicates an increase of plasma NEAC, whereas a negative SMD indicates a decrease in plasma NEAC. Terms a, b or c was used when the same author measured plasma NEAC with different methods. (A colour version of this figure can be found online at http://www.journals.cambridge.org/bjn).
Effect of oxidative stress-related risk factors on plasma non-enzymatic antioxidant capacity
As described in Table 3, participants enrolled in chronic studies were divided into healthy subjects (no detectable exposure to oxidative stress-related risk factors) and subjects exposed to oxidative stress-related risk factors. For the beverage category, an effect on plasma NEAC was clearly detected in the risk factor category (SMD 0·765, 95 % CI 0·310, 1·220; P< 0·001), whereas no changes in plasma NEAC were observed in healthy subjects (SMD 0·177, 95 % CI − 0·154, 0·508; P< 0·296). On the contrary, in the food group (chocolate, fruit, vegetables and plant food-based dietary patterns), a significant increase in plasma NEAC in both categories was observed (P< 0·001 for healthy and OSRRF subjects). However, the efficiency of dietary supplementation was much higher in subjects characterised by risk factors (SMD 1·253, 95 % CI 0·685, 1·820) than in healthy subjects (SMD 0·502, 95 % CI 0·235, 0·769; P< 0·001). Finally, when all the available data (beverages and foods) were grouped, we observed that the increase in plasma NEAC was three times higher in subjects with risk factors (SMD 0·937, 95 % CI 0·592, 1·281; P< 0·001) when compared with healthy subjects (SMD 0·367, 95 % CI 0·162, 0·572; P< 0·001).
Table 3 Chronic dietary supplementation studies with beverages and solid foods containing antioxidants and plasma non-enzymatic antioxidant capacity in human subjects: effect of oxidative stress-related risk factors (OSRRF) (Number of interventions and 95 % confidence intervals)
SMD, standardised mean difference; ↔ , unchanged; ↑ , increased.
* Test for the overall effect.
Effect of methodology on plasma non-enzymatic antioxidant capacity
The influence of methodology on the response of plasma NEAC after dietary supplementation with antioxidant-rich plant foods and beverages is described in Table 4. When methodologies based on the hydrogen atom transfer reaction were analysed, a positive response was detected in TRAP and ORAC, both in acute (SMD 1·113, 95 % CI 0·486, 1·741; P= 0·001; SMD 1·308, 95 % CI 0·704, 1·911; P< 0·001) and chronic studies (SMD 0·574, 95 % CI 0·078, 1·071; P= 0·023; SMD 0·653, 95 % CI 0·304, 1·001; P= 0·001). When single-electron transfer-based methods were tested, FRAP was increased both in acute (SMD 1·059, 95 % CI 0·612, 1·505; P< 0·001) and chronic periods (SMD 0·366, 95 % CI 0·076, 0·656; P= 0·013). On the contrary, the TEAC assay was unable to detect changes in plasma NEAC after dietary supplementation with vegetable foods (acute studies: SMD 0·08, 95 % CI − 0·660, 0·819; P= 0·819 and chronic studies: SMD 0·318, 95 % CI − 0·075, 0·711; P= 0·113).
Table 4 Chronic supplementation studies with beverages and solid foods containing antioxidants and plasma non-enzymatic antioxidant capacity in human subjects: effect of methodology (Number of interventions, participants and 95 % confidence intervals)
SMD, standardised mean difference; TRAP, total radical-trapping antioxidant parameter; ↑ , increased; FRAP, ferric-reducing antioxidant potential; ORAC, oxygen radical absorbance capacity; TEAC, Trolox equivalent antioxidant capacity; ↔ , unchanged.
* Test for the overall effect.
Discussion
The present meta-analysis shows that the ingestion of antioxidants from most vegetable foods and beverages differently modulates plasma NEAC in human subjects. The antioxidant effect is more prominent in subjects exposed to cardiovascular risk factors or affected by oxidative stress-related pathologies.
Analysing the plant-derived beverage category, we noted that red wine, black tea and green tea displayed an excellent antioxidant effect, in both acute and chronic studies. In agreement with epidemiological evidence, studies analysing the antioxidant potential of red wine, included in the present meta-analysis, used a moderate amount of red wine (about 250 ml), suggesting that the protective role of moderate red wine intake could be linked to its ability to enhance the body's antioxidant system. On the other hand, fruit juice failed to display any antioxidant action in chronic trials, differently from single-time ingestion where an antioxidant effect is detected. The heterogeneity of fruit juice composition could be one of the reasons for the variability in plasma NEAC following chronic consumption. Moreover, fruit juices rich in fructose or enriched with high-fructose maize syrup, inducing free radical production, could decrease the antioxidant properties of fruit juice(Reference Lim, Mietus-Snyder and Valente20).
The plant-derived food category, such as chocolate, fruits and vegetables, showed a clear antioxidant response after acute ingestion, whereas only vegetables were able to increase plasma antioxidant capacity after chronic intervention trials.
Despite more evidence being needed, variability of fruit composition and presence of potential oxidative stress inducers, such as fructose, might be the basis of the lack of effect displayed by fruit and fruit juices. Chocolate, despite the high content in flavonols and antioxidants, failed to display any effect in chronic studies. However, if we analyse the chronic studies, we observe that in Fraga's(Reference Fraga, Actis-Goretta and Ottaviani21) study, milk chocolate, losing antioxidant properties in vivo, was tested. Moreover, subjects enrolled in five chronic intervention trials were characterised by a lack of risk factors for oxidative stress, which is a factor able to affect the in vivo antioxidant response of foods(Reference Serafini, Villano and Spera7). In order to determine whether cocoa could represent a valid source of antioxidant ingredients, more studies on people exposed to oxidative stress risk factors are needed.
Epidemiological data linked the protective role of plant foods on several diseases with their ability to ameliorate the body's redox network. In line with evidence suggested in these studies, we reported a strong increase in plasma NEAC in dietary patterns rich in the plant food category.
Considering the overall effects, we observed a difference in terms of NEAC responses on the basis of the study design. In particular, when compared with chronic designs, plasma NEAC is promptly increased after vegetable food ingestion in acute studies. Acute ingestion trials rely on the importance of having a short experimental window free from variables such as physical activity, diet and homeostasis that might affect NEAC in chronic studies. However, as for all the kinetics, once a peak of effect is reached, between 30 and 60 min for beverages and after 2 h for solid foods, plasma levels turn back to baseline levels in few hours. Acute ingestion models allow us to draw relatively fast conclusions about the ‘potentiality’ of the food to display an antioxidant action in human subjects. It might be considered a useful pilot study before a chronic trial, still representing the golden standard for confirming the effect observed in an acute trial. On the other hand, the increase in plasma NEAC detected in chronic studies suggests a positive association between endogenous and dietary antioxidants. In chronic studies, we face a physiological mechanism of homeostatic control, aimed to tune an endogenous antioxidant network. More specifically, it might be possible that in a ‘healthy condition’ such as the absence of specific risk factors for oxidative stress, the body does not require an upload of dietary antioxidants. In order to test this hypothesis, we divided all chronic intervention studies on the basis of the ‘healthy status’ of the subjects.
Interestingly, and in agreement with previous evidence(Reference Serafini, Villano and Spera7), we observed a strong antioxidant response in the ‘OSRRF’ category that is three times higher than the ‘healthy’ category. In particular, plant-derived beverages (tea, fruit juices and red wine) are unable to increase plasma NEAC in healthy subjects, whereas plant-derived foods (chocolate, fruit, vegetables and dietary patterns based on plant-derived food) enhance plasma NEAC both in the healthy and OSRRF groups, with a double effect size in the last category. In this view, when chronic intervention studies were conducted on healthy subjects characterised by the lack of specific OSRRF, the body's antioxidant defences might better cope with free radical formation, preventing the need for dietary antioxidants. On the contrary, an increased intake of dietary antioxidant compounds could be identified as a helpful strategy to potentiate endogenous redox defences if pro-oxidant conditions are present.
In order to evaluate NEAC, several methodological approaches were performed so as to give birth to a major obstacle in understanding the pros and cons of NEAC as a biomarker for assessing antioxidant capacity in vivo. Based on hydrogen atom transfer (TRAP and ORAC) or single-electron transfer (FRAP and TEAC), these methods could provide information related to the redox network in the plasma compartment after antioxidant ingestion(Reference Serafini, Villano and Spera7, Reference Bartosz9). Our data revealed that all methods, except TEAC, showed a similar trend following plant-derived food and beverage consumption. In particular, TRAP and ORAC displayed a double response in plasma NEAC in acute studies when compared with chronic studies, whereas the FRAP method showed a three times higher antioxidant response in acute studies than in chronic studies. On the contrary, the TEAC assay did not reveal any change in plasma NEAC neither in acute nor in chronic trials. The ORAC and TRAP assays measure the antioxidant inhibition of peroxyl radical-induced oxidation and therefore reflect classical radical chain-breaking antioxidant activity by hydrogen atom transfer. On the other hand, the TEAC assay is based on the ability of antioxidants in scavenging the radical cation 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid). Biases of this method are related to plasma dilution and to non-physiological conditions of analysis that strongly reduce its sensitivity to detect plasma NEAC modulation in biological fluids. In fact, the peroxidase process does not mimic radical formation in vivo and the assay uses a non-physiological radical. Although the methods used revealed a good reproducibility in measuring plasma NEAC, among all the methods to detect the hydrogen atom transfer reaction, we suggest to use the TRAP and/or ORAC assays, whereas FRAP could be more sensitive than the TEAC assay (single-electron transfer-based methods) to assess plasma NEAC. Finally, the present results highlight the importance of assessing plasma non-enzymatic antioxidant defences through a battery of methodologies furnishing complementary information on different modalities of action of dietary antioxidants(Reference Lettieri-Barbato, Villaño and Beheydt22, Reference Villaño, Lettieri-Barbato and Guadagni23).
Although several meta-analyses have shown a decreased risk of oxidative stress-related disease after vegetable food consumption(Reference Desch, Schmidt and Kobler24–Reference Aune, Lau and Chan26), there are no meta-analyses measuring the impact of vegetable foods on plasma antioxidant markers in human subjects. In the present meta-analysis, we evaluated 122 intervention studies with 2890 analysed subjects, so that the precision and power of the analysis is increased. Moreover, about 86 % of the included trials were performed with more than ten participants; this highlights the validity of the present results(Reference Bjelakovic, Nikolova and Gluud6).
However, the present meta-analysis has some limitations. First, the reviewed studies in the present meta-analysis are extremely heterogeneous, as they have different study designs. Indeed, in the meta-analysis, about 62·5 % of the total intervention studies were not randomised and controlled, whereas about 46·6 % of the included studies had a randomised and controlled design. Finally, about 37·5 % of all the intervention trials were randomised and controlled with a longitudinal study design.
Second, due to the difficulties in planning an appropriate placebo, clear evidence about the antioxidant potential of plant-derived foods and beverages is still far to come. However, on the basis of the above-described considerations, in order to evaluate the risk of overestimating intervention effects, all trials were stratified according to the study design (experimental and quasi-experimental design), and both EXP and Q-EXP studies have been processed together only after evaluating that their direction effects were similar(Reference Borenstein, Hedges and Higgins13).
In conclusion, our data suggest that ingestion of plant foods and beverages increases plasma antioxidant potential in human subjects. We also showed that the antioxidant effect of plant foods is most conspicuous in subjects exposed to OSRRF, compared with those who are not exposed. These findings suggest the importance of considering exposure to oxidative stress as an important variable in deriving the benefits from the antioxidant action of plant foods. Accordingly, the present findings suggest the importance of introducing the concept of the ‘redox necessity for a redox effectiveness’ of antioxidants, beneficial when the body is under stress and not in all phases of life, if a plant food-based diet is routinely followed. Further chronic intervention studies involving subjects sharing a common need for dietary antioxidants are warranted to unravel the relative contributions of dietary and endogenous antioxidants.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0007114513000263
Acknowledgements
D. L.-B. designed the study and directed its implementation, including quality assurance and control. F. T. contributed to the analysis of the data and revised the article. A. S. contributed to the interpretation of the data and to the writing and revising of the manuscript. G. M. conducted the literature review and contributed to the writing and revising of the manuscript. M. S. helped in supervising the project, designing the study's analytic strategy and writing and revising of the manuscript. No conflict of interest exists for each author. No funding was received.
