For the last couple of decades we have been bombarded with messages to eat more fruit and vegetables. One of the main arguments for this, and indeed the entire motivation behind the “5-a-day” campaign that started in the early 90’s, is that fruit and vegetables protect against cancer. But do they really?
The European Prospective Investigation in to Cancer and Nutrition, otherwise known as the EPIC study (I guess “EPICN” didn’t sound quite as, well, epic) was a large cohort study that looked in to this question. A cohort study is a type of observational study that recruits people and then follows them over time to see what happens to them. The study was funded by the European Commission and the International Agency for Research on Cancer.
The results were published in the Journal of the National Cancer Institute in 2010. The study followed roughly 500,000 people in ten European countries. Participants filled in a questionnaire about what they ate when they were recruited in to the study, and they were then followed over time to see if they developed cancer, with a median length of follow-up of nine years. Intake of fruit and vegetable at the beginning of the study was then compared with risk of developing cancer over the following years, to see if there was any correlation.
Many readers will probably immediately notice one big problem with the study. They’ve interviewed people at one time about their fruit and vegetable intake, and then simply assumed that this stays stable over time. If a lot of people had a low intake of fruit and vegetables at the beginning of the study, and then a year in to the study switched to a high intake, that could seriously mess up the results. So the study builds to a large extent on the assumption that people overall don’t change their fruit and vegetable intake very much over the course of a decade. Is that assumption correct? I have no idea. This problem is very common in observational studies of diet.
A second issue stems from the fact that this was an observational study. Observational studies always have one big problem, and that is confounding. Confounding is caused by the fact that people are not randomized in to a high-fruit-and-vegetable and a low-fruit-and-vegetable group. They’re getting to choose for themselves which group they’re in. Since we’ve all been told for decades that fruit and vegetables are healthy, people who care about their health will generally have a higher intake than people who don’t care. They will also exercise more, smoke less, drink less alcohol, and so on.
If a difference in cancer incidence is found in this study, which of all those different variables is actually causing the difference? The authors of the study have attempted to correct for the most obvious potential confounders, like smoking, alcohol intake, and exercise habits. They have also corrected for educational level, and in women they corrected for menopausal status, age at which menstruation started, and contraceptive use (factors that correlate strongly with risk of breast cancer, which is by far the most common cancer in women). They also initially corrected for intake of red meat, processed meat, and fibre, but then decided not to bother when they realized that these things had zero effect on the results (as we’ve discussed previously, red meat consumption does not increase your risk of cancer).
In other words, they attempted to correct for a bunch of possible confounders. Of course, that leaves about five billion other confounders that they didn’t correct for. That’s the problem with observational studies. No matter how hard you try to correct for confounding variables, you’re never going to catch them all. That’s why observational studies cannot generally be used to prove the existence of a cause-and-effect relationship. The best they can do is suggest that such a relationship exists (in the language of medicine, we say that they are “hypothesis generating”). As a general rule, anything less than a halving or doubling of risk in observational studies should be assumed to be entirely due to confounding, at least until someone proves otherwise.
Let’s get to the results.
Average intake of fruit and vegetables in the study was 335 grams per day. The lowest intake was seen in Sweden, where average intake was 231 grams per day, and the highest was seen in Spain, where average intake was 511 grams per day. Intake of fruit and vegetables correlated positively with female gender, higher education, increased physical activity, lower alcohol intake and lower rates of smoking. This is as we would expect – people who care about their health also eat more fruit and vegetables.
Over the course of the nine years of follow-up, 6,4% of participants developed cancer. Now to the interesting part. The group with the highest intake of fruit and vegetables (more than 650 grams per day) had an 11% lower relative risk of cancer than the group with the lowest intake (less than 230 grams per day). Overall, the researchers found that an increase in consumption of fruit and vegetables of 200 grams per day (equivalent to two and a half servings) was correlated with a 3% reduction in the relative risk of getting cancer.
If we for one second assume that there is a real cause and effect relationship here, and that this reduction isn’t due to confounding, that would mean that someone who increases their intake of fruit and vegetables by two and a half servings per day, perhaps to get up to the recommended “five-a-day”, would decrease their risk of getting cancer over the coming nine years from 6,4% to 6,2%. In other words, the absolute risk reduction over nine years is 0,2%. That’s an effect size that makes even statins seem impressive.
Even if we take the more extreme example of someone who goes from eating no fruit and vegetables at all to eating eight servings per day, the absolute reduction in risk of cancer over nine years would still only be 0,7% (if you’re confused about the difference between absolute risk and relative risk, I recommend you read this).
And that’s assuming there is a causal relationship. As I think I’ve already made clear, there is no reason to think that an effect this small in an observational study is due to a cause-and-effect relationship. In fact, considering the enormous number of possible confounders, it’s almost certainly due to confounding. In other words, the best evidence that exists at this point in time suggests that an increased consumption of fruit and vegetables won’t decrease your risk of cancer.