Ultra-Processed Food: The One Controlled Trial

Matched for calories, sugar, fat, sodium and fibre, people still ate about 500 kcal a day more on the ultra-processed diet.

What this is

Most nutrition evidence is observational, and observational evidence about ultra-processed food is easy to dismiss as confounded. Which makes the NIH inpatient trial unusually valuable: 20 adults living in a metabolic ward, randomised to two weeks of ultra-processed and two weeks of unprocessed meals, matched for presented calories, energy density, macronutrients, sugar, sodium and fibre, eating as much as they wanted.

They ate roughly 500 kcal a day more on the ultra-processed diet, and gained about 0.9 kg over the fortnight, losing a similar amount on the unprocessed diet. Small, short, and a genuinely controlled test of the central claim.

What the evidence says

  • Ultra-processed diets increase spontaneous calorie intake

    Gold· human data

    The NIH inpatient randomised crossover trial, with meals matched on the usual nutrient suspects. Twenty people over four weeks — the design is strong, the sample is small.

  • UPF intake is associated with worse health outcomes

    Silver· human data

    Large cohorts consistently associate high UPF intake with cardiometabolic disease and mortality. Observational, with substantial confounding by income, smoking and overall diet quality.

  • Eating rate and energy density are part of the mechanism

    Silver· human data

    People ate faster on the ultra-processed diet, and softer, energy-dense food is easier to over-consume. Plausible and partially supported.

  • The UPF classification is a precise tool

    Unproven

    NOVA groups wholemeal supermarket bread with confectionery, and the category is defined by processing rather than nutrition. It's useful at population level and blunt at the shopping-basket level.

  • All processed food is harmful

    Not supported

    Freezing, tinning, pasteurising and fermenting are processing, and much of it improves safety, cost and nutrition. The trial tested a category, not the concept of processing.

What this means in practice

  • Judge by what a food contributes: protein, fibre, micronutrients, satiety. That's more useful than its NOVA group.
  • The practical version of the trial's finding: minimally processed food is harder to overeat. Build meals around it where you can.
  • Keep the useful ultra-processed items — tinned pulses, frozen vegetables, plain yoghurt, protein powder if it helps you hit a target.
  • Watch out for the softer, faster-eaten, energy-dense items specifically — that's where the intake difference showed up.
  • Don't turn this into a purity contest. Cost and time are real.

When to see a doctor

  • Escalating food rules, anxiety about ingredients, or avoiding social eating.
  • Cutting food groups to the point of restricted variety or nutritional gaps.

The honest summary

The best controlled test we have found people eat several hundred extra calories a day on an ultra-processed diet even when the nutrients are matched. That's a real finding worth acting on loosely — and the classification is too blunt to justify anyone's purity rules.

Work it out

See also

Frequently asked questions

Is ultra-processed food bad for you?
The controlled evidence says people spontaneously eat substantially more of it, which matters for weight. The disease associations come from cohort studies with real confounding. Overall: worth reducing, not worth panicking about.
Is protein powder ultra-processed?
By the NOVA classification, generally yes — which illustrates the category's bluntness. Whether it's useful depends on whether it helps you hit a protein target you'd otherwise miss.
How much did people overeat in the trial?
About 500 kcal a day more on the ultra-processed diet than the unprocessed one, gaining roughly 0.9 kg over two weeks, despite meals matched for calories presented, sugar, fat, sodium and fibre.

Sources

Written and reviewed by Mathew Beale, MSc Biotechnology, University of Reading.

Last reviewed: