News · Brain & Mental Health
Ultra-processed food tied to more depression in over-70s
A Molecular Psychiatry analysis of 7,220 older adults found high ultra-processed food intake tracked with more depressive symptoms, and genetic risk made no difference either way.
Based on a peer-reviewed cohort analysis in Molecular Psychiatry
- Researchers at Deakin University used the ASPREE cohort of 7,220 adults aged 70 and over, followed for a median 5.7 years, publishing in Molecular Psychiatry.
- Rather than a standard cohort analysis, they used target trial emulation, a method designed to mimic a randomized trial using observational data.
- High ultra-processed food consumption was associated with a 12% higher risk of depressive symptoms (RR, 1.12; 95% CI, 1.03-1.21).
- An anti-inflammatory diet was associated with 7% lower risk, though the interval just touched the null (RR, 0.93; 95% CI, 0.86-1.00).
- Genetic predisposition to depression, measured by polygenic score, did not modify the association.
- Neither did low-grade inflammation, measured by C-reactive protein, which weakens the most popular proposed mechanism.
- Depressive symptoms were measured by questionnaire, not clinical diagnosis, and diet was self-reported once.
The idea that diet shapes mood has a credibility problem, and it is not the biology. It is that miserable people eat differently, which makes every cheerful finding suspect. A study in Molecular Psychiatry took that objection seriously enough to redesign around it.
Researchers followed 7,220 Australians aged over 70 and found that high ultra-processed food consumption was associated with a higher risk of depressive symptoms. Two things that were supposed to explain the link turned out not to.
Fixing the method first
Previous work on food and mood has been repeatedly undercut by the same three problems, which the authors name directly: reverse causality, unmeasured confounding, and selection bias remained important limitations.
So rather than run another association analysis, they emulated a target trial of dietary interventions using data from a large Australian cohort of older adults. Target trial emulation forces you to specify the randomized experiment you wish you could run, then rebuild it from observational records. It is stricter than a conventional cohort analysis, and it fails more honestly.
What was measured
An ultra-processed food index and an anti-inflammatory diet measure were extracted from a food frequency questionnaire to score diet quality. Depressive symptoms were assessed annually with a Center for Epidemiologic Studies-Depression 10-item score of 8 or above.
Two additional layers made this more than a diet study. A polygenic score was derived using the latest Psychiatric Genomics Consortium data for major depression, and systemic inflammation was assessed from a routine blood marker.
A total of 7220 participants, 52.7% female, aged 70 and over, were followed for a median of 5.7 years.
The results, in two halves
The diet findings ran in the expected direction. High ultra-processed food consumption was associated with a higher risk of depressive symptoms, around 12% higher. An anti-inflammatory diet was associated with lower depressive symptoms, about 7% lower, though that interval reached the point of no effect.
The second half is where it gets interesting, because both of the study’s headline hypotheses failed. Genetic predisposition or low-grade inflammation did not modify the observed associations.
Why the failures matter more than the finding
Take the genetic result first. A common assumption is that lifestyle matters most for those already at inherited risk. Here it did not: the association looked much the same whether or not someone carried a high polygenic score for depression.
The inflammation result cuts deeper. Low-grade inflammation did not modify the observed associations either. The dominant explanation for why processed food might affect mood is that it drives low-grade inflammation, which affects the brain. If that were the whole mechanism, the link should be strongest in people with the most inflammation. It was not.
That leaves the association standing and its explanation open. Nutrient displacement, gut microbial changes and blood sugar variability all remain candidates, and none was tested here.
What to hold back
Depressive symptoms came from a ten-item questionnaire, not a clinical assessment, so this measures symptom burden rather than diagnosed depression. Diet came from a food frequency questionnaire, which people fill in imperfectly.
And target trial emulation, for all its discipline, is still observational. It narrows the space for reverse causation; it does not close it. Five and a half years is not long enough to be sure which came first in a condition that develops slowly.
The authors’ own conclusion is appropriately hedged, noting that higher diet quality is associated with a lower risk of depressive symptoms, independent of genetic predisposition or low-grade inflammation, which may support dietary interventions as a modifiable lifestyle strategy for mental health.
For a 70-year-old, the practical version is unchanged by the mechanism failing: less packaged food, more of everything else, and no reason to think your genes let you off.
People also ask
What is target trial emulation?
It is a way of analyzing observational data that forces the researchers to specify, in advance, the randomized trial they wish they could run: who would be eligible, what the treatment and comparison are, when follow-up starts. They then reconstruct that design from the cohort. It does not eliminate confounding, but it removes several common biases, including the one where people are classified by something that happened after follow-up began.
How big is a 12% higher risk?
Modest. It means the high ultra-processed food group had roughly 12% more depressive symptoms over the follow-up than the low group, not that 12% of them became depressed. On a common outcome in an older population that is a meaningful population-level signal and a small individual one. The anti-inflammatory diet result was weaker still, with an interval reaching 1.00, the point of no effect.
If your genes predispose you to depression, does diet matter more?
Apparently not, and the researchers looked specifically. Genetic predisposition, captured by a polygenic score built from the largest available depression genetics data, did not modify the observed associations. The practical reading is that the diet signal looked similar regardless of inherited risk, so nobody appears exempt and nobody appears especially vulnerable on this evidence.
Wasn't inflammation supposed to be the mechanism?
That is the popular theory, and this study does not support it. Low-grade inflammation, measured with high-sensitivity C-reactive protein, also did not modify the associations. If ultra-processed food raised depression risk purely by driving inflammation, you would expect the link to be stronger in people with higher inflammation. It was not, which points elsewhere, perhaps to nutrient displacement or the gut.
Could depression be changing the diet rather than the reverse?
It is the obvious rival explanation, and the reason the authors used target trial emulation at all: they name reverse causality, unmeasured confounding, and selection bias as the limitations that dogged earlier work. The method reduces those problems without abolishing them. Low mood reliably changes what people cook and buy, and five years of follow-up cannot fully separate the two directions.
References
- Mengist B, Davoodian N, Lotfaliany M, et al. Diet quality and depressive symptoms in older adults, assessing the effect modification by genetic predisposition and low-grade inflammation: a target trial emulation. Molecular Psychiatry (2026).
- National Institute of Mental Health. Older Adults and Mental Health.