Diet and dying: people who ate more fruit and veg were less likely to die

shopping basket with fruit and veg

Yum!  There are few morsels more delectable to an Elf’s palate than a meta-analysis of observational studies of the effects of good diet.

Because diet is a long-term investment, it can be hard to design prospective randomised trials to investigate the benefits. Often we need to look at observational studies. Although they can’t establish causation, they can point a pretty sturdy finger in its direction, especially when they’re combined in a systematic review.

Another benefit is that they can more readily look at really important long-term clinical outcomes, such as mortality, and can be less reliant on surrogate end points such as blood pressure, cholesterol or glycaemic control.

Of particular interest in this new review is the promise to investigate the “dose-response” relationship between fruit and veg and mortality risk.

Clinical question

The clinical question wasn’t explicitly stated in this review. It seems to be “what is the association between dietary intake of fruit and vegetables and the risk of dying from cancer, cardiovascular disease or indeed any cause?”

It’s always a little worrying when a review doesn’t make its scope completely clear up front using the “patients, interventions, outcomes” format. However, on we go.

The review investigated the relationship between

The review investigated the association between eating fruit and veg and the risk of dying from cancer, cardiovascular disease and other causes.


The systematic review took in MEDLINE, EMBASE and the Cochrane Library databases.  The reviewers also looked up the references cited by the papers that they found.

From their search results, they identified relevant studies using blind, independent assessment of study eligibility and methodological quality.

From the eligible studies, they extracted data about participants’ consumption of fruit and vegetables and the risk of mortality.

The reviewers performed a mind-boggling routine of statistical gymnastics to obtain estimates of the association and dose-response relationship between diet and death.


The reviewers found 16 studies involving 833,234 patients. Within the studies, there were reports of 56,423 deaths in total (11,512 from cardiovascular disease and 16,817 from cancer).

The studies varied in size and duration, ranging from 501 to over 450,000 participants, with 4.6 to 26 years of follow-up.

The key finding of the analysis was:

The pooled hazard ratio of all cause mortality was 0.95 (95% CI 0.92 to 0.98) per increment of one serving of fruit and vegetables a day, with significant heterogeneity (P<0.001, I2=82%).

When they plotted the “dose-response” relationship, they found that the benefits tended to tail off above 5 servings per day.

This evidence, therefore, supports the current recommendation to take five servings of fruit or vegetables per day.

This evidence

This study adds further evidence to the five a day public health message.

Our Elf concludes

It’s worth pointing out that there appeared to be benefits from lesser amounts.

It can take some time to wean an elf’s palate off its customary diet of deep fried toadstools and tonic honeydew. This research shows that positive gains may be made through a gradual adjustment of diet, and we don’t have to terrify everyone with broccoli for breakfast, lunch and tea.

There are two concerns with the reliability of the research, though.

  1. Firstly, when the reviewers performed a sensitivity analysis for study quality, they found that poorer quality studies tended to find a greater benefit. This might indicate bias in favour of diet. However, it’s worth pointing out that the higher quality studies still showed a statistically significant if slightly reduced benefit for all-cause mortality.
  2. Secondly, they detected “significant heterogeneity”, meaning that the findings of the individual studies were inconsistent.  This might  indicate the presence of biased studies in the review, but it might also mean that the studies are looking at different things and it wasn’t valid to combine all the different studies into one overall estimate.

So we do need to sound a quiet note of caution here.

Heterogeneity is good for your diet, but not for your systematic review.

Heterogeneity is good for your diet, but not for your systematic review.


  • It could be that the reviewers didn’t find all of the evidence with this search strategy as reported. However they did follow up on references cited by the studies they found, which is a good thing.
  • It’s not clear whether there was a prospective publication of the review protocol.
  • The reviewers made some assumptions to convert categorical reports of levels of fruit and veg consumption into numerical ones for the dose-response analysis.


Wang X, Ouyang Y, Liu J, Zhu M,Zhao G, Bao W & Hu F. Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies. BMJ 2014;349:g4490