Generations of parents have encouraged their children to eat their vegetables. There is an almost intuitive assumption that plenty of vegetables and fruit is an important part a healthy diet. In the 1956 Four Food Group Plan, the cornerstone of dietary guidance in the USA for almost forty years, the target was at least four daily portions from the fruit and vegetable group. In 1990, the World Health Organisation (WHO) recommended that we should all eat at least 400g of fruit and vegetables each day i.e. five 80g portions. This has led to five-a-day health promotion campaigns in several countries including a UK campaign in 2003.
These campaigns promoting fruit and vegetable consumption were prompted by many epidemiological studies indicating that people who eat higher amounts of fruit and vegetables have lower overall death rates than those with low intakes. High fruit and vegetable consumption is associated with increased life expectancy and reduced death rates from cardiovascular diseases. There is also a perception that high fruit and vegetable intake reduces cancer risk, often linked with their high level of so-called antioxidants. The reduction in cancer mortality is actually less clear than cardiovascular disease. Cancer is a blanket term that covers many diseases, so lumping them all together may mask a beneficial effect upon certain cancers especially as some non-dietary factors like smoking, alcohol, excessive sunlight exposure, asbestos, radiation and other environmental exposures are major causes of some cancers.
A study in the UK published in 2014 ( Oyebode et al, 2014 ) led to many headlines questioning whether the five portion a day recommendation should be increased to seven e.g. on the BBC news web-site (1/4/2014) or even 10 a day in the Daily Telegraph (31/03/2014) . I posted a short piece in response to these newspaper headlines in April 2014. Similar headlines suggesting our five a day recommendations should be increased to 10-a-day have recently appeared again (e.g. on the BBC news web-site (23/02/2017) ) in response to a very large meta-analysis published by a group at Imperial College London (Aune et al, 2017 ).
A summary of evidence underpinning the 5-a-day recommendation
I have selected a few recent studies that are representative of the consensus of the current literature but bear in mind that the WHO felt that even in 1990 there was already sufficient evidence to formally recommend 400g/day (five 80g portions).
Fruits and vegetables have a number of positive characteristics that contribute to their healthy image:
- They are sources of several essential nutrients like vitamin C, folate (B9), riboflavin (B2), beta-carotene (vitamin A) and potassium
- They are almost all naturally very low in fat and calories
- They are a major source of dietary fibre including soluble fibre
- Eating lots of fruits and vegetables may displace other foods with less healthy images
- They provide large numbers of plant chemicals (phytochemicals) for which many health claims are made, albeit with very limited evidence in almost all cases
- They are the major sources of antioxidants although clinical trials suggest that antioxidant supplements are more likely to do harm than good for affluent populations.
A recent meta-analysis of sixteen cohort studies with well over 800,000 participants who were followed for between 5 and 25 years (45,000 total deaths) was published in the BMJ (Wang et al, 2014 ). In this study, higher consumption of fruit and vegetables was associated with a dose dependent reduction in risk of death. Each extra portion of fruit and vegetables was associated with a 5% decrease in total mortality but only for the first five portions. A similar stepwise reduction was also noted when they looked specifically at cardiovascular mortality risk but there was no significant association with risk of death from cancer.
An earlier meta-analysis (Dauchet et al, 2006 ) published in the Journal of Nutrition looked specifically at cardiovascular disease. They amalgamated cohort studies with over 200,000 subjects and a total of over 5000 “cardiovascular events” and found a 4% reduction in event risk for each extra portion of fruit and vegetables and an even more marked reduction in fatal events.
Both of the studies discussed above are meta-analyses or weighted amalgamations of several cohort studies. In cohort studies subjects are assessed and then followed to see if there is a relationship between initial measurements i.e. fruit and vegetable consumption and outcome e.g. risk of death, heart attack or cancer diagnosis. Cohort studies are regarded as the most powerful of the epidemiological methods but, of themselves, can only show association between fruit and vegetable consumption and health outcome and cannot prove cause and effect. Certain features of these studies increase the likelihood of the association being due to cause and effect:
- The effect is relatively strong
- The effect is graded i.e. increasing benefit with increasing dose
- The results are consistent with other studies of several types
- There are plausible reasons/mechanisms to explain these results
- The results suggest a specific effect of fruit and vegetables because the effect remained after “correction” for other influences; most of the individual studies attempted to correct for obvious confounding effects e.g. age, fatness, smoking, alcohol and some for activity level.
The European Prospective Investigation into Cancer and Nutrition (EPIC) is one of the largest cohort studies ever conducted with around half a million subjects spread across many centres in several different European countries. This study was designed specifically to try to identify dietary factors associated with increased or decreased risk of cancers. Boffetta et al, 2010 published an analysis of the relationship between fruit and vegetable consumption and overall cancer risk after about 9 years of follow-up. They reported only a very small decrease in overall cancer risk associated with higher fruit and vegetable consumption. This decreased risk was so small that the authors suggested that because of the uncertainty involved in correcting for confounding variables, like tobacco and alcohol use, this association should be interpreted cautiously. This report was greeted with the online headline (8/04/2010) Fruit and veg offer “limited protection” against cancer by the BBC. Data from this large European cohort has also confirmed a 4-5% reduction in cardiovascular deaths for each of the first 4 or 5 daily portions (Crowe et al 2011 ) .
Up to 2014 there thus seemed to be clear and consistent evidence that high fruit and vegetable intake was associated with reduced all cause and cardiovascular mortality and good grounds for believing that this is a causal association i.e. high consumption of fruit and vegetables increases life expectancy and reduces risk of cardiovascular disease. These past results seem to suggest that the effect is graded but tends to tail off at around 400g per day i.e. consistent with five-a-day recommendations. Evidence that high fruit and vegetable consumption will reduce cancer risk was less clear and any reduction seems to be much smaller.
The main findings of the 2014 paper
Oyebode and her colleagues at UCL used data on a representative sample of around 65,000 English adults aged over 35 years that was collected as part of the Health Survey for England surveillance programme. Individuals within this study were linked to mortality outcome data so that their deaths and their causes could matched up with the diet and lifestyle data. They found that mortality from all causes, from cardiovascular disease and from cancer decreased with increased fruit and vegetable consumption.
At the extreme, roughly 8.2% of those eating less than 1 daily portion died within 8 years but only 4.1% of those in the highest category (over 7-a-day) died. Of course, several other factors would have contributed to these differences and the authors attempted to correct for some of these other factors; in their “fully adjusted model” they attempted to correct for age, sex, smoking status, social class, educational attainment, body mass index (fatness), physical activity and alcohol consumption. This is the most difficult and problematical part of any epidemiological study because there is no statistical magic wand that can precisely correct for these other variables especially when the information on them amounts to little more than a crude categorisation. Nevertheless, even after their correction process there remained a pronounced and step-wise reduction in mortality with increased fruit and vegetable consumption as shown in figure 1. The correction has narrowed the gap between the two extremes but a substantial difference remains. The authors tried various different correction processes e.g. excluding deaths within the first year of the study which could be due to existing illness. These different correction processes did not substantially alter the pattern shown in figure 1.
The main findings of the 2017 study by Dagfinn Aune and her colleagues
This is a large complex study amounting to 28 pages, 7 large tables and about thirty individual graphs and diagrams. The study involved a meta-analysis of 95 separate cohort studies involving 2 million subjects from around the world looking at the relationship between fruit and vegetable consumption and all-cause mortality, cancer and cardiovascular diseases. They were able to show a strong dose dependent association between high fruit and vegetable consumption and reduced risk of death from all causes combined, cardiovascular diseases (both stroke and coronary disease) and for total cancers. They report that these graded effects were observed up to 800g/day (i.e. 10 portions) except for cancer where a plateau was reached at 600g/day (7.5 portions). However, their graph of fruit and vegetables and all-cause mortality shows quite a steep and fairly linear decline in mortality up to 5 portions per day with a distinct flattening off once this point is reached and only modest extra gains seen when going from 5 to 10 portions per day; this is similar to that seen in figure 1. A similar flattening is seen in the graph of fruit and vegetable intake and cardiovascular disease. This flattening is even more pronounced in the fruit and vegetables and cancer graph and the overall decline is noticeably less than in the previous two graphs.
This paper adds considerable extra statistical power to previous findings because of the higher number of included studies and the 2 million total subjects but it does not really alter previous findings. Because of this extra statistical power, it allows the authors to explore more detailed analyses which are not discussed here e.g. between different categories within the fruit and vegetable category and different diseases under the cardiovascular diseases umbrella.
Conclusions about the benefits of fruit and vegetable consumption
Recent studies confirm that increasing fruit and vegetable consumption is associated with a progressive decrease in all-cause mortality, cardiovascular disease mortality and total incidence of cardiovascular disease (fatal and non-fatal); these findings seem to be robust and newer studies have added further weight to these existing conclusions. The picture with respect to cancer is not quite so clear cut; the 2014 meta-analysis of Wang et al found no appreciable reduction in cancer deaths associated with increased fruit and vegetable consumption. Data from the EPIC study found only a relatively small and weak association between cancer incidence and fruit and vegetable consumption although it did confirm the 4-5% per portion reduction in cardiovascular disease. This is a very powerful and well-designed study with over half a million participants in a single cohort so its results merit high weighting.
The huge 2017 meta-analysis of Aune and her colleagues did report an association between cancer incidence and fruit and vegetable consumption but it was less than for cardiovascular disease and all-cause mortality and seemed to tail off at a lower intake than total or cardiovascular mortality.
A major unanswered question is whether these associations represent cause and effect i.e. does increasing fruit and vegetable consumption cause these improvements in health and mortality risk. This question is technically unanswerable with the epidemiological methods that are available. It is possible that the differences in mortality and disease incidence associated with increased fruit and vegetable consumption are due to other factors (confounding variables) that are linked to both fruit and vegetable consumption and the mortality/disease risk outcomes. As a simpler example, if one found a strong association between high alcohol consumption and lung cancer risk this could be because people who consume high amounts of alcohol are also more likely to be exposed to cigarette smoke, the major cause of lung cancer. One would have much more confidence that the association was causal if it remained substantial after one had accurately corrected for the level of exposure to cigarette smoke. In an earlier section, I listed a number of characteristics of the association between fruit and vegetable consumption and mortality/disease risk that made it much more likely that the association was causal and even back in 1990 the WHO was confident enough of this causal association to make the 5-a-day recommendation. Most health scientists, including myself, accept that this association is almost certainly causal although it is still possible that high fruit and vegetable consumption is simply a marker for being very health conscious and having a generally healthy diet and lifestyle.
Table 1, uses data from Oyebode et al 2014 to illustrate some of the difficulties of “correcting” for confounding variables. The under 1-a-day category are twice as likely to die during the course of the study but they also contain more of the very elderly category, many more males, many more less educated people, many more regular smokers, more obese people, more physically inactive people and more heavy drinkers. Interestingly, the under 1-a-day category also has more of the youngest age group. The statistical analysis tries to eliminate the effects of these other influences but the categorisation of these other parameters is very crude or limited and so there are inherent difficulties in this correction process. Information on other potential confounding variables e.g. other dietary differences was simply not collected as part of the Health Survey for England and so could not be corrected for. Fat and particularly saturated fat intake is an obvious, unconsidered confounding variable.
Total calorie intake was also not assessed and one might expect fruit and vegetable consumption to tend to rise with increasing total food intake. On average, males would be expected to eat more than females and the youngest might be expected to eat more than the oldest but more males and more of the youngest are in the under 1-a-day category. Those undertaking vigorous activity might be expected to eat more than those who are inactive so this might offer a partial explanation for the decreasing numbers in the less than 1-a-day category in the more active groups.
The relationship between fruit and vegetable consumption and total mortality shown earlier in figure 1 has been corrected by the authors for all of the available other variables show in table 1. The fact that even after correction the differences in mortality risk remain substantial gives more credibility and confidence to the belief that this association is causal. However, the smaller the differences become, the less confident one would be e.g. the difference between 5 to 7-a-day and 7+-a-day is very small so how confident can we be that this is a real causally-related difference? The other differences between the extremes, the under 1-a-day and the 7+-a-day are very substantial and the 7+-a-day group contains the smallest number of people.
Table 1 Percentages of different categories of people with different characteristics in the highest and lowest fruit and vegetable consuming categories
|Characteristic||Under 1 portion/day (%)||Over 7 portions/day (%)|
|Never regular smoker||11.0||10.9|
|BMI under 20||21.0||7.8|
|BMI 30+ (obese)||15.7||9.1|
|Moderate (within guidelines)||11.0||10.4|
|High (over double guidelines)||21.8||4.2|
Do these new studies really justify increasing the 5-a-day recommendations?
Even if we take the values in figure 1 at face value they suggest that most of the benefit of increasing fruit and vegetable consumption would be from increasing the consumption of the currently low consumers. Moving from the under 1-a-day to the 3 to 5-a-day category was associated with a reduction in total mortality of about a quarter, with a further fall of 6% when moving from 3 to 5 to 5 to 7-a-day and only another 3% when moving from 5 to 7-a-day to 7+-a-day. Small differences obtained using current epidemiological methods should be treated with extreme caution and yet some media reports have suggested increasing current recommendations to 7-a-day on the basis of this study and some even suggested 10-a-day despite the fact that there was not even a 10-a-day category!
Figure 2 has been constructed from the data that Oyebode et al (2014) extracted from the Health Survey for England. It shows how the English adult population was distributed between the various fruit and vegetable categories. This means that 25 years after the WHO 5-a-day recommendation and over a decade after the start of the UK campaign, only a quarter of adults are reaching five-a-day and almost half (46.1%) are eating less than three portions per day. One would expect that the bulk of the 7+-a-day category would be concentrated in the 7 to 10-a-day range leaving just a very few per cent currently in the 10+-a-day category. Average number of portions of fruit and vegetables per person was 3.8 per day (made up of 2.3 fruit and 1.5 vegetables).
Taken together with the consensus of the scientific literature, what figures 1 and 2 suggest is that getting the 75% of the population eating less than 5-a-day up to that level would be confidently expected to yield substantial population health benefits. Focusing upon raising still further the fruit and vegetable intake of those meeting current guidelines offers only a relatively low probability of producing a small extra benefit.
Is 10-a-day a sensible or even responsible recommendation?
There seems to be no evidence to support a 10-a-day recommendation from the work of Oyebode et al and only limited evidence from the huge meta-analysis of Aune et al of a relatively small extra benefit for minority of the population who are already meeting current guidelines. There are a number of important issues that need to be considered before we raise current recommendations.
- Would a 7 or 10-a-day target seem so unattainable for the bulk of the population that their motivation to raise their low current intake would be diminished?
- Most of the research has focused upon middle-aged or even elderly adults but how appropriate would a very bulky 10-a-day diet be for rapidly growing children and adolescents?
- Fruit and vegetables are not a very cost effective way of supplying calories to meet the energy needs of poorer families. Each fruit and vegetable calorie probably costs around three times as much as calories supplied by the other food groups. Would 7-a-day or even 10-a-day be financially realistic for the poorer members of society especially if they want to include some of the more appealing and expensive fruit and vegetable items in their new, much higher intakes?
- What would be the impact of mass implementation of new 7 or even 10-a-day recommendations? To get the bulk of the population over 7-a-day would probably involve increasing population consumption of fruit and vegetables by two and a half times and to get them over 10-a-day maybe a four-fold increase. At least in the short term this would lead to severe shortages and much higher prices. Would many poorer people be effectively priced out of the market for these already relatively expensive products (in terms of calories per penny)?
- What would be the environmental impact of trying to meet the extra demand for these bulky and calorie-light products? Fruit and vegetables supply only around 6% of current calorie intake so trebling their intake would only have a modest effect in reducing the need for other foods; it would have a huge impact on the weight and volume of food needed to adequately feed the population. Most of these extra fruit and vegetables would almost certainly be imported with a resultant increase in the “food miles” needed to feed the UK. In producer countries, including some relatively underdeveloped countries, much more land, water, fuel and agricultural chemicals would be needed to generate this level of extra supply.
- Would the nutritional status of people in poorer countries be adversely affected by this massive increase in demand for fruit and vegetables from richer countries? Would the health and nutrition of poorer populations be sacrificed to allow minor and speculative extra health benefits for the richest populations?
Of course no one really believes that we are going to get most of the population of the UK over 10-a-day in the foreseeable future (even reaching 5-a-day seems a remote possibility). Therefore none of these potential hazards associated with a massive, several fold increase in fruit and vegetable demand is a likely to be a real problem. However is it responsible to make recommendations that if widely implemented would have enormous negative impact simply because one is safe in the knowledge that very few people will follow them anyway?