Please note that a modified version of what follows was published as a Rapid Response in the BMJ Online
Or in other words, does fruit and vegetable consumption reduce all-cause mortality? This is the subject of a paper published yesterday in the BMJ (British Medical Journal), which resulted in an article on the BBC News Website by Health Editor Helen Briggs entitled “Fruit and veg: More than five-a-day ‘no effect’. While the study authors conclude that “this meta-analysis provides further evidence that a higher consumption of fruit and vegetables is associated with a lower risk of all cause mortality, particularly cardiovascular mortality”, somehow this has within the space of 24 hours been translated in the media to “there is no need to eat more than five portions a day”.
Of course this new story falls hot on the heels of recent suggestions that there may be further health benefit to be obtained from increasing fruit and vegetable consumption to seven portions a day. A backlash was always to be expected, as many still struggle to achieve five. For the majority, health benefit “up to five-a-day” remains a positive message, but for those in the vegetarian/vegan/raw communities who routinely consume much more, this may sow a seed of doubt that all their efforts to consume 10 or 15 portions, or even up to 80-100% of their total calories from fresh fruits and vegetables might be a waste of time. If this applies to you, fear not, and read on.
In this study, fruit and vegetable consumption was divided into six categories:
- One portion per day
- Two portions per day
- Three portions per day
- Four portions per day
- Five portions per day
- Six or more portions per day
For each category something called a pooled hazard ratio was calculated. Hazard ratios express the risk of a certain outcome (in this case death during the period of follow-up) for a particular level of consumption in comparison to no consumption. So a hazard ratio (HR) of 0.92 (for one portion per day) indicates that eating one portion of fruit or vegetables per day reduces your risk of dying (during the follow-up period) that is 92% of the risk you would have had if you had not consumed any.
Here are the hazard ratios:
- One portion per day HR = 0.92
- Two portions per day HR = 0.85
- Three portions per day HR = 0.79
- Four portions per day HR = 0.76
- Five portions per day HR = 0.74
- Six or more portions per day HR = 0.74
Taken by themselves, it does look as if each sequential portion added reduces your risk of dying by less than the last one (something the authors call a curvilinear association). If true, actually this doesn’t seem that unreasonable. Your first portion of the day does you the most good, and the second does you good but not quite as much, etc. Increasing you consumption from fifteen to sixteen portions may do you only a marginal amount of good, but that does not mean it does you no good at all, and this is still quite different from suggesting it does you harm.
Anyway the story does not end here. The authors also report 95% confidence intervals for these figures. The confidence intervals are a range within which we can be 95% confident that the true value lies (since the quoted figure is subject to sampling error: if you repeated this study 100 times you would get 100 different results). The larger the sample size (how many people they studied), the narrower the confidence intervals, because bigger studies produce more reliable results.
Here are the confidence intervals:
- One portion per day HR = 0.92 CI = 0.90-0.95
- Two portions per day HR = 0.85 CI = 0.81-0.90
- Three portions per day HR = 0.79 CI = 0.73-0.86
- Four portions per day HR = 0.76 CI = 0.69-0.83
- Five portions per day HR = 0.74 CI = 0.66-0.82
- Six or more portions per day HR = 0.74 CI = 0.65-0.82
If the confidence intervals for any two groups being compared do not overlap, then we can be 95% confident that the observed difference is true. For example, if we compare eating one and three portions of fruit and vegetables per day, we can see that the confidence intervals do not overlap. However, if we compare eating three and six or more portions, there is significant overlap, so we cannot say with any confidence that there is a real difference. This may in part be because the confidence intervals become wider with each subsequent portion of fruit and vegetables.
Here, in brackets, are the widths of the confidence intervals:
- One portion per day HR = 0.92 CI = 0.90-0.95 (0.05)
- Two portions per day HR = 0.85 CI = 0.81-0.90 (0.09)
- Three portions per day HR = 0.79 CI = 0.73-0.86 (0.13)
- Four portions per day HR = 0.76 CI = 0.69-0.83 (0.14)
- Five portions per day HR = 0.74 CI = 0.66-0.82 (0.16)
- Six or more portions per day HR = 0.74 CI = 0.65-0.82 (0.17)
You can see that the width for six or more portions is more than three times the width for one portion. Why is this? Almost certainly it is because in the seven studies assessed, there were fewer and fewer subjects in each subsequent category (i.e. there were lots of people who consumed one portion per day, and not very many who consumed six or more). This data is not provided in the original article, but they do provide a table (Table 2) at the end that gives categories of consumption for each of the studies included in the meta-analysis. For those that report consumption in terms of servings per day as opposed to grams per day or times per week (two of the seven studies), the maximum consumption categories were 5.9 servings per day of fruit and 5.2 servings of vegetables for one study, and 5+ servings for the other. Interestingly the study that reported a maximum of 5.9 servings of fruit and 5.2 servings of vegetables was the JACC study, conducted in Japan (where they advocate seventeen portions of fruits and vegetables per day). The other study was the Baltimore Longitudinal Study of Aging. Unfortunately the JACC study alone does not give sufficient power to this meta-analysis to draw firm conclusions as to the benefits of increasing fruit and vegetable consumption at higher levels of intake. The two studies reporting consumption in times per week gave maximum consumption categories (for fruits and vegetables separately) of 4 to 7 and 6 to 7 times per week, indicating relatively low consumption levels in their study populations.
And finally, to the person on the Facebook thread who asked:
“has any study been made on fully raw vegans? To actually reveal the TRUE RESULTS of eating raw fruits and veggies? They never need any medical care! There are so many of them, young and old, nobody makes any study on them.”
I’m afraid the answer is still no, I don’t think so. Wouldn’t that be great?! What we need is a prospective matched cohort study of raw vegans. Now that might really tell us something new!