The short answer
In one Dutch birth cohort, children who ate the most ultra-processed food at age three had a larger rise in BMI z-score by age ten to eleven than children who ate the least. The difference between the highest and lowest quarters was 0.22 BMI z-score units, with a 95 per cent confidence interval of 0.07 to 0.37. A second analysis designed to test direction suggested that early diet predicted later BMI more clearly than early BMI predicted later diet. That is an association from an observational study, not proof that a single food causes weight gain.
The study, in one paragraph
Yang, Navis and Corpeleijn at University Medical Centre Groningen analysed two Dutch cohorts in a paper published in BMC Medicine on 28 September 2026. The you
nger group was GECKO Drenthe, a birth cohort with 1,091 children whose diet was assessed at age three with a validated food frequency questionnaire. The older group was Lifelines, with 2,970 children aged eight to twelve at dietary assessment. In both groups, foods were categorised with the NOVA classification, and height and weight measured by trained nurses were used to calculate age and sex standardised BMI z-scores. In GECKO, children were followed to age ten or eleven. In Lifelines, follow-up covered ages nine to seventeen.
What counts as ultra-processed food?
The researchers used the NOVA system, which groups foods by the extent and purpose of industrial processing rather than by a single nutrient such as sugar or fat. In this study, food items reported on the questionnaire were placed into NOVA categories, allowing the team to estimate how much of each child’s total intake, by weight, came from ultra-processed products.
The amounts were high. In GECKO at age three, the median ultra-processed intake was 744 grams per day, accounting for about 52 per cent of total food intake by weight. In Lifelines at ages eight to twelve, the median was 908 grams per day, about 50 per cent of total intake by weight. In plain terms, roughly half of what these children ate, measured by weight, fell into the ultra-processed category.
That definition matters for how you read the result. NOVA is a broad grouping. It puts together products that differ in ingredients, portion sizes and nutritional profile. The study therefore tests a dietary pattern dominated by ultra-processed products, not the effect of one biscuit, one drink or one ready meal in isolation.
What happened between age three and age ten to eleven?
The clearest result came from the youngest children. In GECKO, children in the highest quarter of ultra-processed intake at age three showed a greater increase in BMI z-score between age three and age ten or eleven than children in the lowest quarter. The reported coefficient was 0.22, with a 95 per cent confidence interval from 0.07 to 0.37.
A BMI z-score is not the same as an adult BMI reading. For children, BMI changes naturally as they grow, so researchers standardise it for age and sex. A z-score tells you where a child sits relative to children of the same age and sex, and whether that relative position is drifting upwards or downwards over time. Here, the highest intake group drifted upwards more.
The authors also categorised overweight and obesity status using International Obesity Task Force criteria, which are designed for international comparison in children. The headline finding in the abstract is about BMI development, though, not a simple count of how many children crossed a threshold.
Why does the direction of the arrow matter?
Parents and clinicians often face a chicken and egg problem. Do children gain more weight because they eat more ultra-processed food, or do children who are already growing faster simply eat more of everything, including ultra-processed food? Observational studies cannot settle that question the way a randomised trial could, but the team tried to get closer to it.
In the GECKO cohort they used cross-lagged panel models, a method that compares relationships over time in both directions. Does diet at an earlier point predict BMI later, after accounting for earlier BMI? And does BMI at an earlier point predict diet later, after accounting for earlier diet? The abstract reports that these analyses indicated a directional association, with early ultra-processed intake predicting later BMI z-score.
That finding is the reason the study is interesting beyond the 0.22 figure. It weakens, but does not remove, the simple reverse explanation that heavier toddlers just ate more processed food. The authors still describe early childhood as a potentially sensitive period, language that signals a hypothesis strengthened by this data rather than a settled fact.
What happened in the older children?
The older Lifelines group did not show the same pattern. Higher ultra-processed intake at ages eight to twelve was associated with lower BMI z-scores at ages nine to seventeen. Compared with the lowest quarter, the coefficients were minus 0.12 for the second quarter, minus 0.20 for the third quarter and minus 0.13 for the highest quarter, with confidence intervals that excluded zero.
The authors are cautious about that inverse result. In the paper’s conclusions they state that the evidence for the adolescent association was less consistent. In other words, you should not read the Lifelines finding as evidence that ultra-processed food protects against weight gain in teenagers. Age differences in reporting, growth, activity, family circumstances and the wider diet could all influence an observational result, and the study design cannot rule out residual confounding.
This split between the two cohorts is also a useful reminder of what developmental timing might mean. A diet pattern at age three, when growth and appetite regulation are developing rapidly, is not necessarily interchangeable with the same pattern reported at age ten. The study suggests timing may matter, and it explicitly calls for further longitudinal research across developmental stages.
What does a 0.22 rise in BMI z-score actually mean?
It is a group average difference in trajectory, not a prediction for an individual child. The researchers compared the change in standardised BMI between the highest and lowest quarters of intake over roughly seven to eight years. A coefficient of 0.22 means the high intake group’s relative BMI position rose by about a fifth of a standard deviation more than the low intake group’s, after the modelling used in the paper.
That does not translate neatly into a fixed number of kilograms for every child, because children’s height, age and sex all affect what a given BMI value means. It also does not mean that moving a child from the highest to the lowest quarter would reverse the change by the same amount. Intervention effects have to be tested in intervention studies. What the figure does show is a measurable upward drift in relative weight status linked to the early diet pattern, with a confidence interval that did not cross zero.
What can this study not prove?
It cannot prove causation. Diet was measured with a food frequency questionnaire, which relies on reporting by parents for young children and is subject to measurement error. Height and weight were measured by trained nurses, which strengthens the outcome data, but dietary reporting remains an estimate.
Even with statistical adjustment, observational nutrition research can leave residual confounding. Families differ in income, education, activity, sleep, neighbourhood food access and many other factors that travel alongside diet. The authors’ own wording reflects those limits. They report prospective associations and a directional signal from the cross-lagged models, and they conclude that further longitudinal research is needed.
It also cannot tell you that every ultra-processed product carries the same risk, or that an occasional processed food will change a child’s trajectory. The exposure studied was the overall share of the diet from ultra-processed products at a very young age, sustained as a pattern long enough to be captured by the questionnaire and related to growth years later.
What should parents take from this?
The most defensible takeaway is about pattern and timing, not panic. If roughly half of a toddler’s intake by weight comes from ultra-processed products, this study links that pattern at age three to a larger rise in relative BMI by late primary school age. Reducing reliance on ultra-processed products in the early years, and replacing part of that share with minimally processed foods, is consistent with the direction the authors highlight for public health research and prevention.
It is not a basis for restrictive dieting in young children, who need enough energy and nutrients to grow. Any concern about a child’s growth trajectory is a conversation for a GP or health visitor who can plot measured height and weight over time, rather than a single BMI calculation at home. For children, the trend across repeated measurements matters more than any one reading.
If you want to check an adult BMI, the formula is weight in kilograms divided by height in metres squared. For children, that adult cut off approach does not apply, which is precisely why this study uses age and sex standardised z-scores and International Obesity Task Force criteria instead.
The bottom line
Do ultra-processed foods cause weight gain in children? This study does not prove cause, but it provides prospective evidence that a diet dominated by ultra-processed products at age three predicts a greater rise in BMI z-score by age ten to eleven. The directional analysis points from early diet towards later BMI, the effect was seen in the youngest cohort rather than consistently in adolescents, and the authors identify early childhood as a potentially sensitive period that needs more longitudinal research. That is a narrower and more honest answer than either processed food is harmless or processed food explains everything.
Frequently asked questions
Do ultra-processed foods cause weight gain in children?
The 2026 BMC Medicine study found an association, not proof of cause. In 1,091 GECKO Drenthe children, the highest quarter of ultra-processed intake at age three had a 0.22 greater increase in BMI z-score by age ten to eleven than the lowest quarter, and cross-lagged models indicated early intake predicted later BMI z-score.
What is a BMI z-score in children?
A BMI z-score expresses a child’s BMI relative to children of the same age and sex. Because BMI changes with growth, the standardised score shows whether a child’s relative weight position is rising, falling or staying similar over time. This study calculated z-scores from height and weight measured by trained nurses.
How much ultra-processed food did the children eat?
At age three in GECKO, the median intake was 744 grams per day, about 52 per cent of total food intake by weight. At ages eight to twelve in Lifelines, the median was 908 grams per day, about 50 per cent by weight.
Did the study find the same result in teenagers?
No. In Lifelines, higher intake at ages eight to twelve was associated with lower BMI z-scores at ages nine to seventeen, but the authors state the evidence for that adolescent association was less consistent and call for further longitudinal research across developmental stages.
Should I put my toddler on a diet based on this study?
No. The study is observational and does not test a diet intervention. It links an overall dietary pattern at age three to later BMI development. Concerns about growth should be discussed with a GP or health visitor using repeated measured height and weight, not restrictive dieting at home.
Sources
- Yang J, Navis G, Corpeleijn E. Ultra-processed food consumption and body mass index development from childhood to adolescence in Dutch cohorts. BMC Medicine. 2026. DOI: 10.1186/s12916-026-05270-4. Published 28 September 2026.
- Study methods as reported in the paper: GECKO Drenthe 1,091 children at age three and Lifelines 2,970 children at ages eight to twelve, validated food frequency questionnaire, NOVA categorisation, nurse measured height and weight, International Obesity Task Force criteria.