Fruit and vegetables and disease risk

Diets rich in fruit and vegetables (FV) are associated with a reduced risk of chronic disease⁽ Reference Boeing, Bechthold and Bub ^{1 )}, and therefore increased consumption of FV is recommended as part of a healthy diet^{( 2 , 3 )}. In this review, the evidence supporting the association of FV with chronic disease risk will be discussed alongside approaches to measure FV intake and strategies to encourage increased FV consumption at the population level.

Measuring fruit and vegetable intake

Methods for measuring FV consumption through traditional dietary assessment methods, such as food diaries, FFQ or 24-h recalls, can be flawed⁽ Reference Kirsh, Peters and Mayne ^{56 –} Reference Tucker, Chen and Vogel ^{58 )}. For example, it is well recognised that self-reported intake can often be inaccurate⁽ Reference Horner, Patterson and Neuhouser ^{59 )} due, for example, to reliance on a participant's memory, the inability of some methods to account for day-to-day variation in intake, or the fact that respondents tend to change their usual eating patterns in order to simplify record keeping and/or to impress the investigator⁽ Reference Lee and Nieman ^{60 ,} Reference Livingstone and Black ^{61 )}. These problems can be further compounded by coding and data entry errors⁽ Reference Lee and Nieman ^{60 ,} Reference Deharven, Charrondiere and Slimani ^{62 )}. Different dietary intake methodologies also tend to agree poorly with each other⁽ Reference Michels, Welch and Luben ^{63 )}, and this has also been shown to lead to variation in the demonstrated associations with disease risk⁽ Reference Bingham, Luben and Welch ^{64 )}.

Given the difficulties associated with the dietary intake methodology outlined earlier, more objective indices of FV intake, such as nutritional biomarkers in biological samples, are therefore of interest. Accurate measurement of the concentrations of compounds found in FV in biological samples would allow the confirmation of low intakes of FV in populations, the evaluation of the association between intake and disease risk, and could also be used as biomarkers of compliance in FV intervention studies. Such biomarkers of intake need to be able to discriminate between differences in intakes⁽ Reference Crews, Alink and Anderson ^{65 ,} Reference Hunter and Willett ^{66 )}, should be non-invasive or minimally invasive⁽ Reference Crews, Alink and Anderson ^{65 ,} Reference Field, Brambilla and Hirsch ^{67 )}, reproducible, easily measured⁽ Reference Stockley ^{68 )} and highly responsive to the intervention being carried out⁽ Reference Crews, Alink and Anderson ^{65 )}. Plasma, serum and urinary biomarkers have been explored as potentially useful indicators of FV intake. For instance, FV are the primary source of carotenoids in the diet and, as carotenoids cannot be synthesised by human subjects, they are considered to be good candidate biomarkers of intake⁽ Reference Jansen, van Kappel and Ocke ^{69 )}. Six carotenoids (α-carotene, β-carotene, β-cryptoxanthin, zeaxanthin, lycopene and lutein) are found in appreciable amounts in human serum⁽ Reference Crews, Alink and Anderson ^{65 ,} Reference Al-Delaimy, Ferrari and Slimani ^{70 )}. Other candidate biomarkers of FV intake include antioxidant vitamins such as vitamin C⁽ Reference Dehghan, Akhtar-Danesh and McMillan ^{71 ,} Reference Padayatty and Levine ^{72 )} and flavonoids⁽ Reference Mennen, Sapinho and Ito ^{73 ,} Reference Mennen, Sapinho and Ito ^{74 )}, including quercetin⁽ Reference McAnlis, McEneny and Pearce ^{75 )}. Some of these compounds have been reliably associated with a particular fruit or vegetable, or a class of FV in observational studies, but less successfully with total FV consumption⁽ Reference Jansen, van Kappel and Ocke ^{69 ,} Reference Jansen, van Kappel and Ocke ^{70 ,} Reference Jansen, van Kappel and Ocke ^{72 ,} Reference Jansen, van Kappel and Ocke ^{76 –} Reference Campbell, Gross and Martini ^{80 )}. This is almost certainly due to the complexity of FV and the large number of bioactive compounds present, but also potentially because of other dietary sources of these compounds.

We have recently conducted a systematic review to examine the usefulness of the main biomarkers of FV intake to act as objective indicators of compliance in dietary intervention studies⁽ Reference Baldrick, Woodside and Elborn ^{81 )}. A comprehensive search of the literature was conducted using six databases, and suitable papers were selected and relevant data extracted (n 96). The papers were categorised into three sub-groups: whole diet interventions; mixed FV interventions; and studies involving individual fruits or vegetables. Overall, the most commonly measured, and most consistently responsive, biomarkers were the carotenoids and vitamin C. Single biomarkers were good predictors of single classes of FV, e.g. quercetin was a good biomarker of onion consumption, and lycopene was a good biomarker of tomato consumption. However, single biomarkers cannot accurately and reliably predict total FV consumption. It therefore remains prudent to measure a panel of biomarkers in FV trials (notably α- and β-carotene, lutein, zeaxanthin, β-cryptoxanthin, lycopene and vitamin C).

It is likely that further research will develop some more novel biomarker approaches to assess FV intake. The association between dietary intake and biomarker status may depend on certain population characteristics, for example BMI⁽ Reference Vioque, Weinbrenner and Asensio ^{82 )}. It must also be remembered that biomarkers will never solely reflect dietary intake of FV, as status will also reflect a number of physiological processes, providing a measure of nutrient availability at the tissue level, and therefore there may always be a role for the more traditional dietary assessment methodologies, perhaps in partnership with biomarker assessment⁽ Reference Pollard, Wild and White ^{83 )}. Recently, for carotenoids and potentially a broader range of markers, combining biomarkers and self-reported measures using regression calibration equations has been suggested to allow more powerful tests of association with disease risk⁽ Reference Freedman, Midthune and Carroll ^{84 ,} Reference Freedman, Kipnis and Schatzkin ^{85 )}. Urinary pH has been suggested as an indicator of FV intake⁽ Reference Welch, Mulligan and Bingham ^{86 )}, while urinary K can act as a biomarker, both of FV intake⁽ Reference Berry, Mulla and Chowienczyk ^{87 )} and diet quality in general⁽ Reference Mente, Irvine and Honey ^{88 )}. The use of Resonance Raman Spectroscopy has also been suggested as a feasible method to measure dermal carotenoids⁽ Reference Mayne, Cartmel and Scarmo ^{89 )}.

A number of studies have suggested that summing individual flavonoids (sum of quercetin, kaempferol, isorhamnetin, tamarixetin, naringenin, hesperetin and phloretin) excreted in 24 h urine samples correlates better with FV intake than individual flavonoid excretion⁽ Reference Nielsen, Freese and Kleemola ^{90 ,} Reference Krogholm, Haraldsdóttir and Knuthsen ^{91 )}. Mennen et al.⁽ Reference Mennen, Sapinho and Ito ^{73 )} found that the same sum correlated with fruit and fruit juice intake, but not total FV intake, but these authors suggested that adding more polyphenols from fruits to the panel might have further improved the accuracy of such a biomarker to take into account the diversity of the fruit sources. It may therefore be possible to combine one or more potential biomarkers of FV intake (including vitamin C and carotenoids) and better predict intake with a combined biomarker approach rather than using single markers or a panel of single markers.

Such a combined biomarker approach outlined earlier automatically leads to the consideration of a much broader range of biomarkers to monitor FV intake, for example, using a metabolite fingerprinting approach. Although there are considerable methodological challenges⁽ Reference Favé, Beckmann and Lloyd ^{92 )}, a number of studies have suggested possible biomarkers of individual FV or classes of FV: prolaine betaine for citrus fruit⁽ Reference Lloyd, Beckmann and Favé ^{93 ,} Reference Heinzmann, Brown and Chan ^{94 )}, novel ascorbate derivatives for broccoli or raspberries⁽ Reference Lloyd, Favé and Beckmann ^{95 )}, sulfonated caffeic acid and sulfonated methyl-epicatechins after raspberry consumption⁽ Reference Lloyd, Favé and Beckmann ^{95 )} and phenyl-acetylglutamine for vegetable intakes⁽ Reference O'Sullivan, Gibney and Brennan ^{96 )}.

Do we need to increase fruit and vegetable consumption? Current population intakes

Before considering how to most effectively increase FV consumption in the general population, it must first be demonstrated that such an increase is necessary. In the National Diet and Nutrition Survey in adults aged 19–64 years in 2004, the mean daily FV intake was 2·7 portions per d for men and 2·9 portions per d for women^{( 97 )}. In the new National Diet and Nutrition Survey adult survey for years 1, 2 and 3 combined (2008–9 to 2010–11), adults (19–64 years) consumed on average 4·1 portions per d and older adults (>65 years) consumed 4·4 portions per d^{( 98 )}. A total of 31% of adults and 37% of older adults met the ‘5-a-day’ recommendation^{( 98 )}. The methodologies differed between these two surveys (moving from a 7 d weighed food diary in 2004 to a 4 d estimated diary and with composite foods now being disaggregated), and therefore direct comparisons between the surveys and conclusions about changes over time are not possible, but it can be concluded with some certainty that the majority of the adult population is still not meeting the 5-a-day recommendation. Similar data exist from Northern Ireland (which was not included in the original National Diet and Nutrition Survey, but is now part of the rolling programme), where in a representative sample of 426 adults >65 years old, the mean intake, measured using a telephone-based recall method, was about four portions per d⁽ Reference Appleton, McGill and Woodside ^{99 )}.

The new National Diet and Nutrition Survey results from the rolling programme are not yet comprehensive enough, with large enough sample sizes, to draw conclusions on the geographical variation in FV intake and how this might impact on disease risk. However, macrosimulation modelling carried out by researchers in Oxford, using data from the Family Food Survey, has suggested that if the average diet in Wales, Scotland and Northern Ireland was improved to that achieved in England, then the mortality gap in these countries could be considerably reduced (reduction in Wales 81% (95% credible interval 62%, 108%); Scotland 40% (33%, 51%); Northern Ireland 81% (67%, 99%)), suggesting considerable variation in diet quality between countries in the UK⁽ Reference Scarborough, Morgan and Webster ^{100 )}. The same research group has used similar modelling techniques to focus on FV intake and concluded that 33 000 deaths per year could be avoided if UK dietary recommendations were met, with more than 15 000 of these avoided deaths being due to increased FV intake⁽ Reference Scarborough, Nnoaham and Clarke ^{101 )}.

Lock et al.⁽ Reference Lock, Pomerleau and Causer ^{102 )} have similarly attempted to estimate the global burden of disease attributable to low consumption of FV. The associations, stratified by fourteen geographical regions, sex and age, were estimated using information on FV consumption in the population and six health outcomes (IHD, stroke, stomach, oesophageal, colorectal and lung cancer). The analysis suggested that 2·6 million deaths worldwide and 31% of CVD (and between 2 and 19% of cancers) may be attributed to inadequate consumption of FV. About 1·8% of the total burden of disease worldwide was estimated to be attributable to inadequate FV consumption, compared with 1·3% for physical activity, 2·3% for overweight and obesity, 2·8% for high cholesterol and 4·1% for tobacco⁽ Reference Lock, Pomerleau and Causer ^{102 )}.

A more recent study by Hall et al.⁽ Reference Hall, Moore and Harper ^{103 )} confirmed a low intake of FV worldwide. The authors found that, of 1 96 373 respondents who resided in fifty-two countries of mostly low and middle income, approximately 78% of both males and females consumed less than five portions of FV daily. FV intakes were assessed by responses to survey questions asked as part of the World Health Survey⁽ Reference Hall, Moore and Harper ^{103 )}.

Encouraging increased fruit and vegetable consumption

Evidence available to date indicates that FV are, overall, beneficial to health and that current intakes in Western populations remain low. Public health specialists are, therefore, faced with the challenge of finding methods for encouraging more people to eat more FV and for sustaining this behaviour in the long term. It is generally agreed that interventions to promote increased FV intake should have a theoretical basis⁽ Reference Craig, Dieppe and Macintyre ^{104 )} and frameworks exist to guide the development of such interventions⁽ Reference Michie, van Stralen and West ^{105 )}. Common behaviour theories and constructs used to effect changes in FV intake have recently been summarised⁽ Reference Thomson and Ravia ^{106 )}.

Two systematic reviews have examined interventions to increase FV intake in adults⁽ Reference Thomson and Ravia ^{106 ,} Reference Pomerleau, Lock and Knai ^{107 )}, with a further two carrying out a similar analysis in children (one focused on interventions within primary schools)⁽ Reference Knai, Pomerleau and Lock ^{108 ,} Reference Delgado-Noguera, Tort and Martínez-Zapata ^{109 )}.

Pomerleau et al. ⁽ Reference Pomerleau, Lock and Knai ^{107 )} published their systematic review in 2005. They included a total of forty-four studies (only excluding acutely ill adults), and showed that small increases in FV intake were possible in healthy adults (of the order of 0·1–1·4 servings per d), and that this could be achieved by a variety of approaches. Positive effects were seen consistently in studies which used face-to-face education or counselling strategies, although interventions delivering information by telephone or via computer were also effective. These successful modes of intervention could collectively be described as peer support, which encompasses a variety of methods including face-to-face or group self-management programmes, peer coaching, the use of community health workers, telephone-based peer support and web- and email-based peer support. In addition, community-based multi-component interventions also produced increases in FV intake. In general, studies were able to demonstrate larger increases in those with pre-existing health disorders.

In a more recent review of behavioural interventions to promote intake of FV in both adults and children, Thomson and Ravia⁽ Reference Thomson and Ravia ^{106 )} only included US-based studies, which resulted in the inclusion of only thirty-four studies. A total of thirty-eight non-US studies were excluded, as were studies that included those with pre-existing disease (n 40). A final condition for inclusion was those studies which specifically reported the theory basis or constructs used when the intervention was designed. Such strict inclusion and exclusion criteria will have resulted in a different profile of studies included compared with the earlier review, but a similar result was demonstrated. An increase in FV intake (+1·13 servings per d) was shown for adults, and +0·39 servings per d for children, and in most studies these increases were statistically significant. However, as the authors note, in most cases these observed increases would not have been big enough to allow the populations included to achieve the recommended intake levels. The studies included in the review employed most commonly the transtheoretical model (with stages of change construct) and social cognitive theory as a basis for the interventions.

The systematic reviews of interventions to promote increased FV intake in adults made a number of recommendations for further research. Pomerleau et al.⁽ Reference Pomerleau, Lock and Knai ^{107 )} considered that the effectiveness of all new interventions had to be tested, particularly in developing countries, where several new programmes had been developed, but without formal evaluation. Any new intervention shown to be effective in a particular setting would have to be retested in a new setting, as factors particular to the local setting may affect the effectiveness of the intervention. The authors suggested that the reports should provide a better description of the methods used in order to allow exact repetition of the interventions, and the reports should include estimates of variability for the selected outcomes in order to allow power calculations. The specific components of intervention should be examined in more depth and how these individual effects differ in different populations. Pomerleau et al. ⁽ Reference Pomerleau, Lock and Knai ^{107 )} also considered that there are further gaps in the literature, suggesting a need to better understand factors influencing FV consumption, including economic, social and environmental factors that affect both food availability and the ability of an individual to make healthy choices, and the barriers to making changes towards increased FV intake.

The recent systematic review by Thomson and Ravia⁽ Reference Thomson and Ravia ^{106 )} was rigorous in terms of selecting only theory-based trials in healthy volunteers, and it applied a number of other exclusions when selecting studies. However, within the studies included, the authors still considered that there are a number of design limitations. Self-reported dietary instruments tended to be used, rather than objective biomarkers; cost, which is a potentially important barrier to increased consumption, was frequently not assessed; and there was a lack of clear linkage between the behaviour theories and constructs and the intervention developed. The authors recommend combining behavioural interventions with social marketing, behavioural economics approaches and technology-based behaviour change models in future studies, and always including both process and outcome evaluation.

Two further systematic reviews have focused on increasing FV intake in children. In 2006, Knai et al.⁽ Reference Knai, Pomerleau and Lock ^{108 )} were able to find fifteen studies that met their inclusion criteria, and ten of these demonstrated significant effects on intake, from +0·3 to +0·99 servings per d. All but one of the fifteen studies were school-based. Multi-component interventions seemed to be the most effective, and the following seemed to be important factors for a successful intervention: an intervention duration of at least 6 months, the complete involvement of the whole school community, for teacher training and curriculum integration to be included, the interventions to include peer leadership and encouragement and involve canteen staff, and finally, to include parents, both at home and in the school.

A second systematic review focusing on children, but including only primary school interventions, was published in 2011⁽ Reference Delgado-Noguera, Tort and Martínez-Zapata ^{109 )}. Nineteen strong and moderate quality studies were included, with a further eight being excluded as being of low quality. Studies were categorised as being game- or computer-based, multi-component, free or subsidised programme, or other studies, and meta-analyses were then conducted. The ‘other studies’ classification included only two studies that could not be meta-analysed; although both of these produced significant increases in FV intake, the free/subsidised category showed no effect of these types of interventions, while computer-based interventions seemed to be effective (standardised mean difference 0·33; 95% CI 0·16, 0·50). Finally, there were eleven multi-component intervention studies, but only seven of these were randomised controlled trials and could be combined in a meta-analysis. Six of the seven individual studies produced an increased intake of FV, but the combined analysis showed no effect of these interventions on intake (standardised mean difference 0·08; 95% CI −0·00, 0·17), although this combined effect did approach significance (P = 0·06). The authors also made suggestions to improve the design of future studies, including larger samples, longer follow-up and clarification of the method of randomisation, while further studies in developing countries are also required.