Counting how much ultra-processed food someone eats has always relied on diet questionnaires and personal honesty. But researchers at the US National Institutes of Health (NIH), the world’s largest medical research body, may have found a more objective way: by reading it in your blood and urine.
In a new study published in PLOS Medicine, scientists have developed what they call a poly-metabolite score, a biomarker-based tool that can estimate how much of a person’s energy comes from ultra-processed foods.
These include packaged snacks, soft drinks, ready-to-eat meals, and other industrially manufactured calorie-dense products and nutrient-poor products. This could be a breakthrough for nutrition research, which has long struggled with the inaccuracies of self-reported diet data.
“The limitations of self-reported diet are well known. With metabolomics, we can get closer to an objective measure of food intake and also understand how diet may be impacting health,” said Dr Erikka Loftfield, lead investigator and researcher at the National Cancer Institute.
The NIH team looked at blood and urine samples for 12 months from two different groups: one observational study of 718 older US adults, and one clinical trial where 20 participants were fed two different diets, one high (80%) and one completely free (0%) of ultra-processed foods, each for two weeks.
The researchers found hundreds of tiny substances in the blood and urine, called metabolites, that were linked to how much ultra-processed food a person ate.
Using machine learning, they created a special score called a poly-metabolite score that could tell how processed a person’s diet was.
These scores clearly showed the difference between when someone was eating mostly processed food and when they weren’t, the study authors noted.
The health risks of diets high in ultra-processed foods such as obesity, type 2 diabetes, and even some cancers are well documented.
But quantifying how much people actually eat is tricky, especially when relying on memory-based food logs or questionnaires. People could forget, under-report, or misjudge portion sizes.