Ketone levels, not weight loss, track symptom change on a ketogenic diet in psychosis

Antipsychotic treatment routinely trades one problem for another: medications that control psychotic symptoms also drive weight gain, insulin resistance, and dyslipidemia, and metabolic syndrome in this population is common enough that clinicians treat it as close to expected. A team at the San Francisco VA Medical Center, UCSF, and Stanford ran the first randomized controlled trial testing whether a ketogenic diet can address that trade-off directly, in 58 outpatients with schizophrenia-spectrum disorders or bipolar-I disorder already taking antipsychotic medication.

For the first month, participants were randomized to a ketogenic diet (KETO, n=28) or diet-as-usual (DAU, n=30); 47 completed this comparison. Compliance was high – 83% of KETO participants maintained ketosis on tested days – and relative to DAU, the KETO group showed larger reductions in weight (Padj<.001), hemoglobin A1c (Padj=.05), and insulin resistance (Padj=.05). This is the part of the trial that was actually randomized, and it is where the controlled comparison ends.

After the 1-month mark, participants from both arms were offered the chance to continue, or start, the ketogenic diet for three more months, on an opt-in basis, with no control group running in parallel. Twenty-five people completed the full four months. Compliance rose further, to 94%. By this point the design is a single-arm, pre-post extension, not a trial – there is no DAU comparison at 4 months, and no randomization decided who stayed on the diet that long. Within this uncontrolled group, positive, negative, and depressive symptoms all improved relative to where participants had started, as did cognitive performance on testing (all P<.001).

The detail worth sitting with is what tracked that improvement. The authors checked whether it was explained by weight loss – the obvious candidate, since shedding weight while on antipsychotics is itself associated with mood benefit in other contexts – and it was not: weight loss did not correlate with the metabolic or symptom changes. What did correlate, in participants who stayed in ketosis, was the blood ketone level itself, tracking improvement in both pre-diabetic markers and depressive symptoms (all P<.001). The authors read this as pointing to ketosis as the mechanism, not the calorie deficit or the pounds lost.

That distinction matters for how the case gets framed to a patient and to the rest of a treatment team. "Lose weight, feel better" and "reach ketosis, feel better" are different instructions, with different adherence demands, different monitoring needs, and a different division of labor between the psychiatrist managing the antipsychotic and whoever is running the diet. A correlation is not a confirmed mechanism, and 25 people finishing an unblinded, non-randomized extension is a small base for a claim about what changed depressive symptoms. But the randomized month at the front of this trial is real, and it already shows metabolic gains large enough to matter for a population whose antipsychotics work against it metabolically from day one.

For practice, the trial does not license switching a psychotic or manic patient onto a ketogenic diet unsupervised, nor does it settle the mechanism question. What it does support is treating a structured ketogenic diet as a metabolic intervention worth discussing with patients on antipsychotics who are already managing weight and glucose problems – ideally with a dietitian or physician tracking ketone levels alongside the usual psychiatric measures, since it was ketone level, not the scale, that lined up with how people felt.