More peas, dear?

As soon as I woke I knew it was going to be one of those days. The first words I heard were from the BBC: researchers had discovered that eating five portions of fruit and vegetables a day (optimistic already) was not enough – we must all eat seven. I was suspicious. I said as much to Mrs Grant as I stumbled towards the kitchen: “residual confounding, socio-economic status”. She ignored me.

Eventually I got round to printing the paper in JECH and read it on the train into town (to hear the wise and insightful Sir David Cox talk at the RSS), with increasing alarm. Every day I see bad stats, of course, but the press coverage for this one makes it potentially very harmful by putting people off even attempting any increased fruit/veg consumption. I’ve no doubt that fruit & veg is good, but I don’t believe there is any decent evidence for 5 or 7 or 10 portions. Now, to be fair, the paper itself expresses some caution. Regular readers will know what’s coming next. UCL’s press release spins it a little bit, mentioning the 7 portions quite a lot. And then the press picked it up and spun it a bit more, into killjoy-docs-say-eat-two-pounds-of-broccoli-or-face-certain-death . It’s kind of nobody’s fault but it went wrong anyway. Like the Iraq War.

Well, that’s my kind words of comfort for the authors of the paper. From here on, it’s going to hurt.

I think there are six major flaws that make this study close to totally uninformative:

  1. Residual confounding, particularly by socio-economic status. SES is measured in the data source, the Health Survey for England (HSE) as the “head of household” having a manual or non-manual job. That’s all there is, and to put that into a regression as a covariate and pretend that SES has been taken out of the equation is sheer nonsense. For me, there is a smoking gun: eating more frozen or tinned fruit & veg is associated with significantly higher hazard of death. That just doesn’t make sense unless it is actually a confounded association. It is so ludicrous that they should have stopped at that point and considered things very carefully.
  2. The fruit & veg consumption in HSE relates to the 24 hours prior to the survey. We know that will balance out over the population, and also that it is no worse than other self-reported measures, but it remains biased.
  3. Several subgroup analyses (but not an exhaustive list) appear and get repeatedly quoted, with little or no rationale whatever for their selection. In every subgroup analysis that appears in the paper, the effect is significant and stronger than in the whole dataset. This may be above board – I don’t know – but it looks very much like cherry-picking.
  4. A linear assumption of hazard ratio for one more portion is assumed at one point, without theoretical justification or reference to the data. Presumably that’s where the idea of ten portions came from; we can just extrapolate a linear trend off the end of the observed data. Eat enough veg and you live forever; eat enough frozen veg and you die immediately.
  5. These are people who have chosen, of their own accord, to eat an awful lot of fruit & veg, or at least say they do. That’s not the same as the  UK population, encouraged and cajoled into eating X portions per day.
  6. Some variables are missing in as many as 62% of the participants. They are included as their own category, which we have known since Rubin (1976) is a very bad idea.

I don’t relish going out on a limb and attacking other people’s work that I’m not intimately acquainted with, but I do so here because it is potentially very harmful. If people are discouraged from even attempting to eat more veg because the bar has been set unrealistically high, that is a massive public health own-goal. Even if points 2-6 turn our to be fine, point 1 certainly isn’t, and that is the worst one.

Here is my own transcript of the BBC Radio 4 Today programme interview, and you will see the researcher is not entirely blameless in encouraging a certain bold interpretation of linear and universal benefit. Success and fame is a corrupting influence (or so I’m told).

JH: We’re used to being told that eating five portions of fruit and veg a day was good for us, and we should try to do it, although apparently two thirds of adults in this country don’t. Now researchers say that the benefits are even greater than we thought, and that eating seven or even more portions a day may have considerable benefits. They link high consumption of fruit and veg to longer life; it’s as simple as that. Lola Oyebode is the lead author of the research which is published today in the Journal of Epidemiology and Community Health, and she joins us now – good morning.
LO: Good morning.
JH: Now, tell us what you found, just putting it as simply as you can.
LO: We looked at the general population of England, and we grouped them by how many portions of fruit and vegetables they ate a day, so we looked at people who ate less than one, one to three, three to five, five to seven, and seven plus, and what we found was in each group, the more fruit and vegetables you ate, the better the benefit to your health, with the group who were eating seven or more portions a day having the lowest risk of mortality.
JH: Ah, well, lowest risk of mortality – but you seem to be suggesting there are wide benefits, that it’s a general prescription for good health.
LO: What we looked at was mortality and we looked at mortality from any cause, death from cancer and death from heart disease and stroke, so those were our outcomes.
JH: So, fruit and vegetables are enemies of the big killers?
LO: That’s right.
JH: Now, when you say seven portions, what do you mean? Seven carrots?
LO: The advice is that you have a variety of fruit and vegetables, so not to eat seven of the same sort.
JH: No…I’m not suggesting that most people would like to eat seven carrots!
LO: Well, actually, I would, but…
JH: Right, well, we’ll keep your personal habits out of it!
LO: A portion is about eighty grams, so that’s one large fruit, or a handful of smaller fruit or veg.
JH: Let’s just talk about the difference between fruit and vegetables. Is there any difference?
LO: Yes, what we found was that vegetables had a greater benefit than equivalent amounts of fruit, but we did still find that fruit gave significant benefit to health.
JH: What is it that causes this benefit to happen?
LO: Well, what we think is that the sugar content in fruit makes it not quite as good as vegetables, and that both fruit and vegetables have lots of micro-nutrients, which are important for the body to work properly, and also lots of fibre, which is good for health.
JH: So, in other words, what you’re saying is that if you want to increase your chances of living a long life rather than an artificially short life, if you cut down on red meat, fatty food, and all the rest of it, and increase your intake of fruit and vegetables, that’s the best thing you can do?
LO: Yes, that’s right.
JH: Does that sum it up?
LO: It does. Well, so we found that all additional portions of fruit and vegetables were of benefit, so even those eating one to three portions were doing significantly better than the people eating less than one portion. So, how ever many you’re eating now, eat more!
JH: And what about the age of the people, and the impact that it has? In other words, I mean if you are seventy, is it still worth increasing the amount of fruit and veg that you eat?
LO: Well, we included seventy year olds in our study – we looked at adults aged thirty five and over in the general population.
JH: So it’s true for everybody – get stuck in. Good news for greengrocers. Thank you very much indeed, Lola Oyebode.

 

Advertisements

2 Comments

Filed under Uncategorized

2 responses to “More peas, dear?

  1. Good post: there are a lot of problems with the way this study has been reported in the media.
    In addition to the ones you mention (and I agree that residual confounding is the most serious), another really huge problem is that the study didn’t actually find a statistically significant difference between those who ate 5 portions a day and those who ate 7 or more, which rather spoils the message that 5 portions a day isn’t enough.
    I’ve written my own blogpost about this paper here:
    http://dianthus.co.uk/should-we-eat-7-portions-of-fruit-and-veg-a-day

  2. Pingback: On methodological terrorism | Robert Grant's stats blog

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s