A new study has audited the literature challenging the definition of ultra-processed foods and found that nine in ten of the critical papers had authors with food-industry ties. The audit classified ties broadly, covering funding, consultancy, employment and roles in industry-linked organisations, and compared the critical papers with the wider research conversation. The result is a map of influence rather than a verdict on any single paper. Its authors argue that when a scientific definition faces concentrated, commercially connected opposition, that fact belongs in the open, because editors, guideline committees and journalists all treat the volume of published criticism as a signal of how unsettled a question is.
An audit of the critical literature
The definition being challenged comes from NOVA, a system that classifies foods according to processing. Group four, the ultra-processed category, includes industrial formulations such as soft drinks, sweets, packaged snacks, instant products and reconstituted meats, typically built from extracted substances and additives with little intact whole food. Researchers adopted NOVA because it predicted patterns that nutrient data alone did not: across many countries, the more of a diet comes from group four, the higher the average intake of energy, free sugars, sodium and fat, and the lower the fibre. Cohort studies have associated such diets with increased incidence of several chronic diseases, while acknowledging that association is not proof of a processing-specific mechanism.
NOVA in plain terms
Critics of the definition make arguments that deserve answers rather than dismissal. They question whether such different products belong in one group, whether classification by processing misleads consumers about individual foods, and whether public policy should target specific nutrients instead. Independent academics have raised versions of all three points. The audit’s finding is about the centre of gravity of the published critique: overwhelmingly, it sits with authors connected to the industry that a strict definition would regulate. That does not refute the arguments, but it changes their interpretation, much as a legal argument is read differently when you learn who retained the lawyer. Provenance is evidence about evidence, and mature fields learn to use it.
Why provenance changes interpretation
The mechanism the authors worry about is subtle and well documented in other sectors. Funded programmes choose questions, frame doubts and sustain a stream of publications that can make a settled area appear open. Systematic reviewers counting papers, rather than weighing independence, can inherit the distortion. Public bodies try to guard against it: the World Health Organization applies conflict screening to guideline panels, and in the US the FDA solicits comment publicly so competing evidence is visible. Even so, audits like this one are rare in nutrition, partly because declarations of interest are scattered and inconsistently formatted, which is why the authors call for standardised, searchable disclosure as a basic research infrastructure.
Implications for journals and guidelines
Journals are the obvious pressure point. Many now require competing-interest statements, but placement and enforcement vary, and a declaration on page nine does little for the reader who meets the paper through a headline. The study implicitly asks editors to treat concentrated industry authorship of a controversy as a fact worth surfacing in editorials and press coverage. Guideline writers, meanwhile, can adopt the practice of sensitivity analysis: does the conclusion change if industry-linked papers are set aside? We applied similar reasoning in this earlier report when weighing competing health claims. Where conclusions survive that test, they are stronger; where they collapse, the field learns something important about what was holding them up.
As with all such audits, precision about limits matters. A tie is not misconduct; industry employs scientists, funds legitimate questions and sometimes publishes results against its own interests. Nor does the audit prove NOVA is the right framework, only that its printed opposition is unrepresentative of researchers as a whole. Nutrition science retains genuine uncertainties, about mechanisms, about thresholds, about how processing interacts with nutrients and eating behaviour. The audit narrows one kind of uncertainty, the social kind: it tells us whose voices have been amplified in this particular argument, and invites us to recalibrate how representative we assumed those voices to be.
Expect the findings to feature wherever labelling and dietary guidance are next debated, because those arenas run on cited literature. For readers, the durable lesson is a reading skill: before absorbing a study’s conclusion about food, find its funding line. It takes a minute and changes the meaning of what follows. The ultra-processed question will ultimately be settled by better data, feeding trials, clearer mechanisms and replications by independent teams. Until then, knowing that nine in ten definitional attacks carried industry ties is not the end of the discussion. It is the context that makes an honest discussion possible.
