The Doctor’s Surgery Theodore Dalrymple
‘Diet books perhaps don’t have the desired effect. There’s always room for one more’
We are badgered all the time with dietary advice and have been badgered for hundreds of years.
In his bestselling Essay of Health and Long Life, published nearly 300 years ago, the physician George Cheyne advised us to stick to white meat and vegetables, and to drink only in moderation. ‘There is nothing more ridiculous,’ he wrote, ‘than to see tender, hysterical and vapourish people, perpetually complaining, and yet perpetually cramming.’
As he himself weighed 32 stone at the time and had difficulty moving around, he knew the problem intimately and from the inside, as it were.
The last time I looked, there were more than 70,000 diet books available on Amazon, suggesting (to me, at least) that the question of what it is right to eat, from the point of view of health, has not yet been settled beyond reasonable doubt; and furthermore that diet books perhaps do not always have their desired effect. There is always room for just one more.
Despite the fluctuating advice – eggs are good; eggs are bad; eggs are good again – we tend to pay attention to dietary advice because, after all, we are what we eat. We want to believe that the existential limits of life are, at least to an extent, under our voluntary control.
Cheyne had nothing but common sense to guide him but, these days, most dietary advice is derived from experiment and above all from epidemiology.
These studies usually involve complex mathematics: and the doctor who passes on the latest advice – for example, that longevity is proportional to a person’s consumption of nuts (according to a paper published not very long ago in the New England Journal of Medicine) – is, in effect, passing on what is for him a kind of rumour. He has read the conclusion and hopes that it follows from the evidence that preceded it.
Alas, sometimes it does not. This may be for any of number of reasons. In 2017, there was a re-analysis of the data in a paper published 35 years earlier in the New England Journal of Medicine that came to a very different conclusion from that published then. Different methods yield different results, even where error or dishonesty are not the explanation.
Recently, a well-known researcher in the field of obesity and overeating (an expanding field, if I may so put it) had many of his papers retracted because of the errors – to put it kindly – discovered in them. Up till then, he had been a celebrated guru on how to persuade whole populations to eat less: for example, by offering them smaller proportions of their frequently rather appalling favourite foods.
Most of the conclusions of his experiments seemed commonsensical enough, which perhaps is why no one bothered to question his flawed statistical methods which did not support his conclusions.
In one of his experiments a soup bowl was surreptitiously refilled as people drank from it, and he found that people drank soup more to empty their bowl than to satisfy their hunger. This sounds right to me.
One of his errors was to mine his data for conclusions. If you generate enough data, something supposedly interesting will emerge. If you try to correlate a disease with hundreds of foods, some of them will be found to correlate with that disease. The correct procedure, however, is to work from a hypothesis to data, not from data to a hypothesis.
Even more important, though: do not treat meals as if they were medical procedures. Turbot is not an antibiotic, nor are artichokes a cure for otosclerosis.