checkedThe Carlat Psychiatry Podcast2026-11-02last verified 2026-11-02

did people in the depression diet trial cut ultra-processed food by 85%?

verdict: misattributed
a real number attached to the wrong thing
“they succeeded in following it, cutting their ultra-processed foods by 85% on average.”

a host, a psychiatrist said on The Carlat Psychiatry Podcast, 2026-11-02, at 15:02. hear it in context →

the episode: Kraepelin on Clancy Part 1 (at 15:02). we link the show's own page; we do not re-host audio.

first, the part that holds

the trial is real and the number is real. in a 2026 pilot crossover trial of adults with major depression, participants' mean score on an ultra-processed food screening questionnaire fell from 10.10 to 1.42, a drop of about 86%. 19 of the 20 randomized participants were analyzed, so it is fair to say most participants followed the diet, at least by their own report.

what we found when we went and looked

the figure matches a 2026 pilot crossover trial in a psychiatry journal (20 randomized, 19 analyzed). there, the mean score on an adapted 14-item screening questionnaire for highly processed food (sq-hpf) fell from 10.10 (sd 1.53) to 1.42 (sd 1.43) after the low ultra-processed diet. that is about an 86% drop. the score is not a measure of how much ultra-processed food people ate, though. it counts how many of 14 food groups a person eats at or above a set frequency, and each group scores either 1 or 0. so an 86% drop in the score is not an 86% drop in intake. the questionnaire's developers published an equation that converts the score into an estimated share of daily food grams. applying it gives a fall from roughly 45% to roughly 13% of grams, about a 71% relative reduction. that is our own extrapolation, and it may not carry over to the adapted version this trial used. the paper never states a percentage, and the authors say the trial had no objective measure of ultra-processed food intake.

assertion by assertion

assertionverdictwhat the record says
participants cut their ultra-processed food intake by 85% on averagemisattributedthe mean sq-hpf score fell from 10.10 to 1.42, about 86%. but sq-hpf is a 0-14 count of food groups eaten at or above a frequency cutoff, with each group scored 1 or 0. it is not a measure of grams, calories or servings. someone can keep eating a food group below the cutoff and score 0 for it. the 85% belongs to the score, not to intake. [2]
participants succeeded in following the low ultra-processed dietconfirmed19 of 20 randomized participants were analyzed, and the self-reported screening score fell sharply (p<.001). the limits: adherence was self-reported only, the trial was open-label, and the mean 'ease of intervention' score was a modest 5.74 (sd 1.85). [1]
the paper itself reports an 85% reductionunverifiablethe full text we read gives the before and after means but no percentage. 85-86% is arithmetic from those means. the 'more than 85%' wording that circulated appears to come from press coverage, not from the paper's text. [1]
translated into estimated share of daily food grams, the reduction is about 71%unverifiablethe developers' calibration equation is hpf (% g/day) = 3.7 x score + 7.6. it gives about 45% before and 13% after, a relative fall of about 71%. this is our estimate. the equation comes from a spanish adult cohort, the trial used an adapted version, and the trial authors did not apply it. [2]

both directions

for the speaker: the trial exists and tested what the speaker described. adults with major depression and high ultra-processed food intake (screening score of 8 or more, the top tertile) switched to a low ultra-processed diet that they shopped for themselves with guidance. the self-reported screening score fell from 10.10 to 1.42 (p<.001), about 86%, so the speaker's 85% is accurate arithmetic on the reported means. retention was high (19 of 20). even after converting the score to estimated intake with the developers' equation, the change is still large, about 71% relative.

against: the 85% describes a screening score, not intake. the sq-hpf scores each of 14 food groups 1 or 0 depending on whether a frequency cutoff is reached (for example, sugary drinks twice a week or more). cutting a food group from three times a week to once a week drops the score for that group to zero without anything close to a 100% cut in that food. the converted estimate is closer to 71% than 85%. the authors also list as a limitation that 'there is no objective measure of UPF, as this study relied on participant self-report'. the trial was open-label, so participants knew which diet phase they were in when they answered the questionnaire. it was a pilot built to test feasibility and acceptability, not to measure exact dietary change.

what this cannot say

the trial is small (19 analyzed), open-label and self-reported, and it was built to test feasibility, not to measure the effect precisely. the pubmed record has no abstract. we read the full text through the journal's guest-access link and could not see the supplement, so we could not check exactly how the questionnaire was adapted or exactly when the post-intervention score was taken in the crossover. our 71% figure applies a calibration equation from a different population to an adapted instrument, so it is an approximation, not a finding of the trial. none of this says anything either way about whether the diet change improved depression, which is a separate question.

sources, each one fetched and read

  1. PubMed record 42776545: Reducing Ultraprocessed Foods From the Diets of Patients With Depression: A Pilot Crossover Trial. JAMA Psychiatry, 2026 What we read there: no abstract available; the plain-language summary describes a pilot crossover trial assessing the feasibility and acceptability of removing ultraprocessed foods from the diets of patients with major depressive disorder. Fetched 2026-11-02 (HTTP 200). https://pubmed.ncbi.nlm.nih.gov/42776545/
  2. Integrative development of a short screening questionnaire of highly processed food consumption (sQ-HPF). International Journal of Behavioral Nutrition and Physical Activity, 2022. PMC8785596 What we read there: 14 food-group items, each scored 1 if consumption meets a frequency threshold (e.g. sugary drinks >=2 times/week) and 0 otherwise, total 0-14; 'HPF consumption (% g/day) = (3.7 x sQ-HPF score) + 7.6.' Fetched 2026-11-02 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC8785596/

speakers are described by role, not named. one sentence is quoted under fair use, with attribution and a link to the original at the timestamp; we do not re-host audio or video. nothing on this page is a reason to start, stop or change a medication. if you have the file behind anything marked unverifiable, the research board is where people can answer. adam