It is a truth universally acknowledged that the latest health research published in a respectable journal must be reported in dramatic and anxiety-inducing language. Some recent studies into ultra-processed foods (like fish fingers, fizzy drinks and ready meals), made for alarming headlines. The Sun, for example, told us that ‘just 4 portions of processed food a day could kill you’, while the Telegraph highlighted that ultra-processed foods ‘could increase the risk of early death by 60 per cent.’
As is often the case, the research being reported is well-designed, thorough and cautious. What’s more, the news coverage is not exactly false. The problem is that the numbers driving the headlines are difficult to interpret, and often seem more frightening than they really are.
Let’s unpack what the evidence means in this case.
The Spanish researchers followed around 20,000 students over 14 years, using questionnaires to estimate how many portions of ultra-processed food they were eating every day. They found that those who ate the most ultra-processed foods had a 62 per cent increased risk of death from any cause, compared with those who ate the least.
If we assume the conclusions are accurate, then a 62 per cent increased risk of death sounds pretty shocking. But what does it mean? Surely our risk of death is already 100 per cent, so a risk of 162 per cent might sound like nonsense.
To make sense of it, let’s imagine two friends – Wanda and Winona. They’re both around 50, the same weight, do the same amount of exercise, have same family histories of disease, but not exactly the same habits or lifestyle. Wanda is keen on instant noodles, fish fingers and fizzy drinks, while Winona prefers steamed vegetables and wholemeal artisan sourdough. Wanda is in the group eating the highest proportion of ultra-processed foods, while Winona is in the lowest.
Each one faces an annual risk of death, whose technical name is their ‘hazard’. The key statistic reported in the Spanish study was a hazard ratio of 1.62. This means that, for two people like Wanda and Winona who are similar apart from their different diets, the one with the risk factor – Wanda – has a 62 per cent increased annual risk of death over the follow-up period (around 14 years in the Spanish study).
It’s tempting to interpret this as meaning that Wanda’s life is going to be 62 per cent shorter that Winona’s, but that’s not what’s going on.
Crucially, the risk of dying in any one year for the average person is pretty low, so a 62 per cent annual increase in risk of death may not be as alarming as it sounds. In the UK, for example, the average annual risk of death for 50-year-old women is around 0.2 per cent, which means you might expect 2 in every 1,000 women like Winona to die each year. A 62 per cent increase on that number gives around a 0.3 per cent annual risk of death, which would be 3 in every 1,000 women like Wanda.
The difference, 1 in 1,000, sounds a lot less alarming than the 62 per cent which made the headlines.
What’s more, it remains perfectly possible that Wanda will live longer than Winona. In fact, there’s some elegant maths based on the hazard ratio that lets us work out the probability that one or the other will die first. It turns out, even with Wanda’s penchant for fizzy drinks and fish fingers, there’s still a 38 per cent chance she will outlive Winona!
So what should a cautious consumer make of all this? Should you change your diet?
A major caveat is that the study only shows an association between consuming ultra-processed foods and a higher risk of death – it doesn’t prove what’s causing it. The Sun’s headline, ‘just 4 portions a day could kill you’ is misleading – it might not actually be the fish fingers and filo pastries to blame. That’s because the group consuming the highest proportion of ultra-processed foods may have been on average poorer, or exercised less, or smoked more than the group consuming the lowest proportion. In other words, it’s possible there’s some other factor that both encourages some people to consume more ultra-processed food and leads to a shorter life.
The researchers always try very carefully to take these ‘confounders’ into account when they do their analyses (and this study did factor in a long list of them), but it’s fiendishly difficult to be certain you have fully removed their influence, and it’s always possible that there’s a lurking factor that you have missed.
That doesn’t mean that the study should be ignored. It definitely counts as evidence that ultra-processed food isn’t good for your health, even if it doesn’t prove a causal relationship. But it may not be a good idea to wait for definitive proof before cutting back on the Turkey Twizzlers.
Listen to our interview with Sir David Spiegelhalter about the power of dodgy statistics on the Science Focus Podcast.
Follow Science Focus on Twitter, Facebook, Instagram and Flipboard