Title, Date of Publication & Journal

Fruit and vegetable consumption and all-cause, cancer and CVD mortality: analysis of Health Survey for England data

Journal of Epidemiology and Community Health

March 31st 2014

Claim supported by evidence?

This paper provides evidence for an association between fruit and vegetable consumption and deaths from all causes, as well as from cancer and cardiovascular disease, with higher consumption associated with lower death rates. The paper does not establish a causal link.

Summary

• This was a cohort study (a type of observational study) which used data from the Health Surveys for England; data from these surveys were linked to routinely collected mortality data.
• This was a large study which collected information on some important confounding factors such as BMI, smoking status, alcohol consumption and physical activity; however, some of these factors were only crudely measured and were not measured for all individuals.
• Participants were asked about their fruit and vegetable consumption in one 24 hour period (the day prior to the survey).

Study Conclusions

The main conclusion is that increased consumption of fruit and vegetables is associated with lower rates of death from all causes, as well as from cancer and cardiovascular disease. The other main conclusion is that vegetables appear to have a stronger association than fruit. These associations remained after taking account of other important factors such as age & sex, BMI, smoking, physical activity, alcohol intake, and social class. However, there is considerable potential for residual confounding – see below under strengths and limitations.

The authors also conclude that consumption of frozen and canned fruit may be associated with higher rates of death. It seems likely that this finding could well be due to the fact that people who eat more frozen and canned fruit may also be eating greater amounts of processed food in general – thus this particular association could be explained by other dietary factors.

The authors acknowledge that they have not established a causal relationship and also acknowledge the other limitations – the potential for bias due to how they have measured fruit and vegetable intake and the potential for unmeasured /residual confounding; however, this was not emphasised enough.

Strengths/Limitations

This is a large and representative study which has collected information on the main confounding factors.

The main limitation is the potential for residual confounding:

• Perhaps most importantly, other dietary factors were not taken into account – those who eat large amounts of fruit and vegetables are likely to have a healthy diet in general (less processed food, for example); therefore, it cannot be established whether the association between increased fruit and vegetable intake and lower rates of death is directly due to the fruit/veg intake or due to having a much healthier diet in general (or indeed, a more healthy lifestyle in general).
• BMI was missing for 14.5% of subjects and physical activity was missing for a large proportion of participants because it was not measured in all surveys, so these factors were not fully taken into account.
• Some variables were not measured accurately – physical activity was only measured as “maximum activity intensity in the past 4 weeks”; alcohol intake was measured as “consumption on the heaviest drinking day in the previous week”; and smoking was also only measured quite crudely (no information of number of cigarettes per day or duration of smoking). This would also mean that these factors were not fully taken into account.
• There is also some potential for unmeasured confounding by socio-economic factors. The results are adjusted for education and social class, classified as manual and non-manual, but there may be other socio-economic factors that are important.

Another limitation is the measurement of fruit and vegetable intake (self-reported, in one 24 hour period) – consumption in one 24 hour period may not reflect usual consumption and some people may be more likely to over-estimate their consumption – if more health-conscious people were more likely to overestimate their fruit and vegetable intake, this would lead to an overestimate of the association.

Glossary

Residual confounding – confounding that has not been adjusted for, either because certain factors have not been measured, or because they have been measured inaccurately.