I wish no harm to the authors of “Fruit and vegetable consumption and all-cause, cancer and CVD mortality: analysis of Health Survey for England data”
J Epidemiol Community Health doi:10.1136/jech-2013-203500
whose recent publication has received approving coverage in the media. We are colleagues at the same godless institution, so they cannot be all bad. But (and you expect nothing less from me) I am not bowled over by their arguments about the benefits of vegetables. You will have caught the general tenor of my criticisms of this sort of work in “Diet is an IQ test” http://drjamesthompson.blogspot.co.uk/2013/11/diet-is-iq-test.html
Following Prof David Colquhoun, I joined him in quoting with approval a paper BMJ 2013;347:f6698 doi: 10.1136/bmj.f6698 (Published 14 November 2013) by John Ioannidis in whose train I ranted thus: “Samples of about 70,000 followed until death (with a proper link to death registers) will be required to identify even a few general patterns in diet which might account for a 5-10% increase in risk. If the studies are to mean anything, IQ, personality, sociological and occupational variables will have to enter the mix, and participants will probably have to be paid to stick to the course, and put up with random visits of inspectors looking in the fridge and the medicine cabinet.”
So imagine my pleasure, or alarm, when this paper turns out to have followed 65,226 persons drawn from a nationally representative sample for 7.7 years and visited them at home to find out what they had eaten yesterday (thus remarkably improving accuracy of their recall) and then linking the respondents to death registers. Rather disarming, isn’t it? The authors seem to have got good data without paying participants or raiding their refrigerators. The authors admit that the main limitation is that measurement of fruit and vegetable intake occurred at only one point in time and relies on self-report. There may be social desirability bias and random error (forgetting) in the recall of fruit and vegetable consumption. However, while short of perfect monitoring, this is a big step forwards. All this is very good, and shows epidemiology at its best.
Undaunted, I moved to the second half of my diatribe “IQ, personality, sociological and occupational variables will have to enter the mix”. Here I have found some things to complain about. Although they included sociological and occupational variables, they did not measure IQ or personality. Frankly, I don’t expect that of epidemiologists, because those measures are often neglected by psychologists anyway.
That aside, the authors carry on doing good things by offering us some old fashioned means and standard deviations, with participants categorized by the number of portions of fruit and vegetables they consume. These are the sorts of simple statistics I can understand. For example, the English eat slightly over 2 portions of fruit, and 1.5 portions of vegetables a day. The propaganda about vegetables has left them relatively unmoved.
Table 1 shows that those who do eat vegetables tend to have non-manual occupations, and the more vegetables they eat, the more likely they are to be in middle class occupations. Do vegetables make you rise in social class? Do bananas telephone? Do efficient compasses misdirect? (Can you spot the origin of the last two questions?)
Similarly, 7-vegetables-a-day types are much more likely to have a university degree than vegetable refusers, who tend to be less educated folk. Also, they are less likely to smoke and are more likely to be physically active. On the other hand, they are just as fat and almost as boozy as everyone else. Those who consumed more fruit and vegetables were generally older, less likely to smoke and more likely to be women, in a non-manual household, with degree level education. Veggie Mummies, yah?
Finally, when it comes to deaths during the study period, here’s the crunch: overall, 6.7% of the sample died during the study period of 7.7 years. The sad fact is that if you are 57 years old you have a 6.7% chance of being dead by the time you are 65 years old. (Or would have been 65, for pedants). Those who eat no vegetables have an 8.2% chance of death, the One to Three vegs a day 7.9%, the Three to Five vegs a day 6.4%, the Five to Seven 5.3% and the Seven Plus vegetables only 4.1%. So, although your chance of dying is relatively low, you can make it even lower by feasting on vegetables.
At first glance, the avid vegetable eaters have half the death rate of the no vegetable eaters. It suggests that vegetables are the cause of the difference. However, it could be that vegetables have nothing to do with it.
In table 2 they offer a “fully adjusted” Model 1: Adjusted for sex, age-group, cigarette smoking and social class; and the even more adjusted Model 2: Adjusted for sex, age-group, cigarette smoking, social class, BMI, education, physical activity and alcohol intake. Of course, as sharp eyed readers you will note that they do not offer a Model 0: adjusting for sex and age, the only things which are truly not controllable by individuals. That is a pity.
In table 2 they use hazard ratios, where eating no vegetables (the highest apparent risk category) is set to 1 and the other conditions lots of vegetables rates as 0.69. This certainly shows the differences with increasing consumption of vegetables, but no longer reveals absolute risk. I prefer table 1. In fact, I would have liked to have seen a correlation matrix. I can read those. I concede that such a matrix would not reveal covariance, but it would allow me to begin to think about the associations between the variables. One or two plots of data would also have helped. In my usual ferreting mode I had a look at the supplementary data.
At about 120 months the fruit effect dies out for some, probably artefactual, reason.
In both these adjusted models and in other variations the effect of vegetable consumption continues to be significant. They go into further detail about vegetables (good) and fruit (slightly less efficacious in keeping you alive) and note that canned fruit seems to slightly increase mortality, probably because of the sugary syrup in which they float.
The authors have bundled together factors that none of us can control like our age and sex, with factors we can control like how long we stay in education and the sort of work we do; with factors we can and probably ought to control like how much we eat and drink. All those different categories are “controlled for”. Some mistake, surely? I can understand the “control” for age. Older people are more likely to die in any time period than younger people. However, if I chose to become a university teacher, why “control” for that choice? I took up that occupation precisely because I thought it would be agreeable, if not well paid, and that I would be highly unlikely to suffer industrial accidents. My choice, plus my ability to get such undemanding light labour against, frankly, rather sparse competition, reveals something about me. It may explain my willingness to follow health advice, or it may simply be that I am a cautious man, minimising my risks in my personal and occupational life. A simple fearfulness of character could explain all the associations.
Consider the adjustments. These are based on the assumption that the cigarette smoking, social class, BMI, education, physical activity and alcohol intake are not related to something which itself has an influence on health. They are seen as imposed external factors which can influence health, rather than a series of behaviours related to an intrinsic factor: system integrity. System integrity is a hypothesized intrinsic characteristic which gives you a good body and a good mind, such that you are healthy and intelligent. This may be related to your genetics and/or a favourable beginning in utero. The one give-away sign of system integrity is fast reaction times to simple stimuli. See the Edinburgh group under Ian Deary for all these findings.
Seen this way, the intelligent live longer and healthier lives not because they are wise, but because they are lucky. They eat vegetables because it seems to be the clever thing to do from a health point of view, and perhaps because they can work out that the need for protein from meat is relatively small, so vegetables are more cost-effective. They may even like the taste of them. They also wear seatbelts, use condoms, brush their teeth, don’t smoke, go for walks, don’t eat or drink too much, study hard, strive to get good jobs and always save money.
The conclusion of this study is that we should eat our vegetables, and 7 portions rather than only 5. Perhaps so. It is still possible simply that bright people live longer, even when they are slightly plump and somewhat boozy. No, my gripe is about the way they have interpreted the findings, and the assumptions which underlie their calculations of hazard ratios. The authors make it clear that “This study has found a strong association, but not necessarily a causal relationship. There are additional unmeasured confounders not included in the analyses, including other aspects of diet.” However, they go on to mention other dietary factors, not the psychological ones.
Vegetables may be good for you. But I have been assured that scientists make a most delicious, nourishing, and wholesome food, whether stewed, roasted, baked, or boiled; and I make no doubt that it will equally serve in a fricassee or a ragout.