People who drink coffee and tea are less likely to suffer dementia, according to a large study published this week. The research is from Harvard and appears in the Journal of the American Medical Association (JAMA), one of my profession’s top outlets. There is every reason to presume its conclusion is wrong.
A total of 131,821 people were followed over a period of up to 43 years. The number of dementia cases was high, giving the study good statistical power. Every few years a dietary questionnaire asked participants how much tea and coffee they drank:
After adjusting for potential confounders … higher caffeinated coffee intake was significantly associated with lower dementia risk … and lower prevalence of subjective cognitive decline.
Caffeine drinkers did better by almost every measure the authors examined.
This new research is noise. So is most medical research
Medical journals ask authors to report their competing interests, so I should say I’m drinking a latte as I write this. The coffee makes me feel more alert, more ready to meet the day. Sharp enough, perhaps, to reject the notion that it saves my ageing brain.
Caffeine speeds up your heart, dilates your bronchi, blocks a neurotransmitter in the brain that signals fatigue, and acts as a diuretic. I drink lattes throughout the day, much to the disgust of my Italophile wife. For her, anything more dilute than an espresso after 11 a.m. is morally suspect. I should also mention that she’s helping run a trial of 20,000 people, testing whether a pill called semaglutide, known in a different form as Ozempic, prevents problems like cardiovascular disease.
As the man who does all the cooking while she sits at her desk late into the night, this is of more than academic interest. Recruiting 20,000 people, randomising them to an intervention or a matched placebo, then following them up for twenty years: this is a major undertaking. Yet drugs like Ozempic are already being used by nearly 2 million people in the UK. Why, then, has she bothered?
The reason my wife is running a trial rather than just tracking some of those 2 million is that observational evidence is treacherous. Coffee and tea drinking aren’t randomly distributed. Those who drink them are different from those who don’t, and not always in ways that can be measured. People who consume caffeine might be healthier and more robust than those who don’t. The JAMA paper knows this. The authors take care not to say that coffee and tea drinking causes a lower risk of dementia, only that there is an association.
‘I used to think that correlation and causation were the same thing,’ a joke has someone saying to their friend, ‘but I’ve been on a course and now I know they’re not.’ Their friend replies that it must have been a useful course. A pause. ‘Maybe.’
If the effect size is big enough, observational evidence can tell you the truth about the world. Smokers have a one in two chance of dying from their habit and lose an average of ten years of healthy life. That’s the scale of difference observational studies can reliably detect, and almost nothing else in medicine comes close.
As effects get smaller, you need randomisation and placebo controls. Observational evidence suggested, for example, that hormone replacement therapy had significant net benefits for health, whereas randomised controlled trials showed otherwise. The apparent benefits were because women who chose to take the supplements were different from women who didn’t. The observational benefits, about twice the magnitude reported in the JAMA caffeine paper, concealed serious harms. Antioxidant vitamins were persistently beneficial in observational studies; when they were properly trialled, the benefits were found not to exist.
This new research is noise. So is most medical research. Most of what’s published is wrong and most of what’s right is still useless. Fundamentally, all research should either seek to generate a new hypothesis or test an existing one. ‘Evidence linking coffee and tea to cognitive health remains inconclusive,’ said this paper. Their hypothesis wasn’t new, and the work the authors undertook wasn’t capable of putting it properly to the test.
Career advancement, funding pressures, and the commercial interests of medical journals create perverse incentives that waste vast amounts of money and human effort. The whole system is enough to drive you to something stronger than coffee. Huge numbers of observational studies have shown that moderate alcohol consumption lengthens your life. People are probably still working on others, despite better quality work having already shown the opposite. I’ll keep drinking wine because the joy is worth the cost. And I’ll persist with my lattes. Not because of delusions about health and longevity, but from the conviction that coffee and tea, like wine, make life richer.
Comments