Ketosis and bipolar disorder: controlled analytic study of online reports

Members of online bipolar disorder forums often report experiences of mood-stabilisation on the ketogenic diet, which has traditionally been used in the treatment of epilepsy. We examined the nature and extent of such reports.

To investigate associations between a ketogenic diet and mood stabilisation among individuals with bipolar disorder.

We undertook an observational analytic study of free-text comments in online forums about mood effects of dietary interventions (ketogenic, omega-3 enriched or vegetarian) classified by a priori categories of change in mood stabilisation in 274 people with bipolar disorder.

There were 141 (85.5%) free-text comments on ketogenic diets that reported a positive impact on mood stabilisation. Reports of significant mood stabilisation or remission of symptoms over a period were substantially higher for a ketogenic diet than for other diets (93/165, 56.4%, 95% CI 48.4-64.1) v. 14/94, 14.9%, 95% CI 8.4-23.7), odds ratio 7.4, 95% CI 3.8-14.1, P < 0.0001), many with detailed reports of the improvements experienced and several lasting for extended periods (months to years). Other reported associations included fewer episodes of depression (in 41.2%, 95% CI 30.6-52.4 of individuals); improved clarity of thought and speech (28.2%, 95% CI 19.0-39.0); increased energy (25.9, 95% CI 17.0-36.5); and weight loss (25.9%, 95% CI 17.0-36.5).

Despite the inherent limitations of the observational data based on self-reports posted online, the association strength and reports of sustained benefit support a hypothesis of a ketogenic diet being associated with beneficial effects on mood stabilisation. Caution should be exercised in interpreting this data until a controlled trial can be carried out to examine this hypothesis. These preliminary observations are generally consistent with a mitochondrial dysfunction component to bipolar disorder aetiology with ketones bypassing a block between glycolysis and the tricarboxylic acid cycle.

None.