The ketogenic diet (KD) has a long-standing place in neurology and has been used for treatment-resistant epilepsy since the 1920s (1). KD consists of a rigidly controlled high-fat, low-protein, and low-carbohydrate diet usually with a 4:1 lipid:non-lipid ratio (fat to protein and carbohydrate ratio) (2). Woodyatt noted that in a normal person in a state of starvation or eating a diet containing low carbohydrate and a high percentage of fat, the ketones acetone, acetoacetate, and beta-hydroxybutyric acid increase (3), and the absence of glucose serves as alternative fuels for the body. KD has been proven an effective treatment in difficult-to-control seizures with its use primarily in children with epilepsy (4, 5), particularly those with epileptic encephalopathies whereby epileptic activity may contribute to severe neurological and cognitive impairments (6). The finding that KD is beneficial for epilepsy was supported by a systematic review (7), meta-analysis (8), and a Cochrane review (9). KD and related diets have been proven useful in pharmacoresistant childhood epilepsy (10).

The mechanism by which KD acts is not clearly understood. However, among the many hypotheses advanced, elevation of brain acetone may account for the efficacy of the diet in epilepsy as it has proven anticonvulsant effects (11). In a variation of the diet, the medium-chain triglyceride (MCT) KD increases plasma levels of decanoic acid, which in vivo has been shown to be anticonvulsant; although the precise mechanism remains unclear (12). In young and adult rats, KD increases concentrations of kynurenic acid (KYNA) in the hippocampus and striatum but not the cortex (13). Elevated levels of KYNA in the cerebrospinal fluid have been demonstrated in patients with schizophrenia (14) and bipolar disorder (15). Pharmacological manipulation of kynurenines is a potential treatment strategy for psychiatric disorders (16).

Currently, there are no international protocols guiding the implementation of the diet, rather dietary recommendations are based on individual treating physician’s advice. Consequently, there exists a need for more standardized protocols for management recommendations for clinical and research use (17). In 2006, a group of 26 pediatric epileptologists and dieticians was convened to create a consensus statement regarding the clinical management of KD. They specified the following absolute contraindications to commencing KD “carnitine deficiency (primary), carnitine palmitoyltransferase (CPT) I or II deficiency, carnitine translocase deficiency, beta-oxidation deficiencies including medium-chain acyl dehydrogenase deficiency (MCAD), long-chain acyl dehydrogenase deficiency (LCAD), short-chain acyl dehydrogenase deficiency (SCAD), long-chain 3-hydroxyacl-CoA deficiency, medium-chain 3-hydroxyacl-CoA deficiency, pyruvate carboxylase deficiency and porphyria. Relative contraindications of KD include the following: inability to maintain adequate nutrition, surgical focus identified by neuroimaging and video EEG monitoring, and parent or caregiver non-compliance” (18). The possible risks of KD must be weighted against its potential value for seizure control or its other benefits (19).

Ketogenic diet has been assessed in a variety of neurological conditions other than epilepsy in both animal models and human trials. In an animal model of amyotrophic lateral sclerosis, SOD1-G93A transgenic mice were fed KD. It was shown that KD led to significant alterations in the clinical manifestation of the disease, specifically a higher motor neuron count in the lumbar spinal cord and preserved motor function (20). KD has also been trialed in rats following controlled cortical impact injury, a model for brain trauma, showing that the diet improves both cognitive and motor functioning (21). In an animal model of multiple sclerosis, the effects of KD on memory impairments and inflammation expressed by experimental autoimmune encephalomyelitis were examined. In mice, it was demonstrated that brain inflammation was associated with impaired spatial learning and memory function, and the administration of KD exerted protective effects against these. The proposed mode of action was through attenuation of the immune response and increased oxidative stress observed in the mice (22).

In humans, KD has been trialed in a number of neurological conditions. In a randomized, double-blind, placebo-controlled, parallel group study in Alzheimer’s disease, an oral ketogenic compound AC-1202 was tested on 152 patients. Regular medications were continued throughout the study. Daily dosing of AC-1202 significantly elevated the levels of beta-hydroxybutyrate 2 h after administration. After 45 and 90 days, patients treated with AC-1202 had significant improvements on the ADAS-Cog scale (23). In a small study of seven patients with Parkinson’s disease, five adhered to KD for 28 days (24). Scores on the Unified Parkinson’s Disease Rating Scale improved in all five as did symptoms such as resting tremor, freezing, balance, gait, mood, and energy levels. These results should be interpreted with caution due to the small sample size, subjective ratings, and the lack of a control group to exclude a placebo effect. The modified Atkins diet (a high-fat, low-carbohydrate diet), which creates a ketotic state was trialed in adolescent patients with chronic daily headaches (25). Due to difficulties adhering to the diet, the study was terminated prematurely. Three participants reported an improvement in headache severity and quality of life; however, they still required pharmacotherapy to manage their condition. In a comprehensive review of KD in diverse neurological conditions, Stafstrom and Rho concluded that there are rich opportunities for further investigation of KD in both the laboratory and clinical practice (26).

The therapeutic advantage of KD has been replicated in animal models of neurological illnesses, and the purported underlying mechanisms include those which improve mitochondrial function (27). Molecular, biochemical, and physiological studies tend to support the assumption that cellular energy status is a determinant for multiple disorders (28). Aberrant energy production has been associated with cancer (29), heart failure (30), aging (31), and neurological conditions such as epilepsy (32) and Alzheimer’s disease (33). The precise pathways by which energy disruption is related to these and other disorders are unknown. There are also strong indications of metabolic pathways involving energy production in the pathophysiology of some mental disorders including bipolar disorder, depression, schizophrenia (34) autism spectrum disorder (ASD) (35), and potentially attention deficit hyperactivity disorder (ADHD) (36). There is also a recognized comorbidity between epilepsy and mental disorders (37), which might indicate some commonality of mechanisms.

Given the degree of interest in KD and neurological conditions, the aim of this narrative review is to examine the effect of the diet in mental disorders. The literature searched in anxiety, depression, bipolar disorder, schizophrenia, ASD, and ADHD.

A comprehensive search of the electronic databases PubMed, PsychINFO, and Scopus for peer-reviewed articles published in English was conducted in the last week of November 2016 and updated in January 2017. Search terms were “bipolar disorder” “manic depress*” “depress*” “schizophren*” “autism” “ASD” “attention deficit hyperactivity disorder” “ADHD” “obsessive compulsive disorder” “OCD” “anxiety” “anxi*” “psychiatry” “mental disorder*” (group 1) AND “ketogenic diet” “ketosis” “ketogenesis” “ketone bodies” “high fat low carbohydrate” “diet” “acetone” “acetoacetic acid” “beta-hydroxybutyric acid” “acetyl-coA” “ketonemia” “ketonuria” “fatty acid metabolism” “hyperketonemia” “fasting” “nutritional ketosis” “acidotic” (group 2). These terms were combined as follows: group 1 AND group 2. In addition, a hand-search of the reference lists of published articles was also conducted, and articles were assessed for their suitability in the review. An initial search was conducted using all the search terms listed above, and abstracts were reviewed by author Emmanuelle C. S. Bostock. Full text publications were retrieved for those that addressed the subject matter.

The results are discussed by mental disorders examining animal and human studies including case reports and studies of patient groups. The search yielded 15 studies that examined KD in mental disorders, specifically anxiety, depression, bipolar disorder, schizophrenia, autism, and, ADHD. These studies included nine animal models, and in humans four case studies and two uncontrolled trials. A summary of results by animal models and human studies are presented in Tables 1 and 2, respectively.

In neurology, KD is an established treatment option for treatment-resistant epilepsy with evidence from a range of studies including controlled trials. By contrast, KD research in humans with mental disorders, though extending over a 50-year period, has received little attention with few studies other than case reports, small sample size open studies, and no controlled trials. Animal studies have been more systematic, investigating mechanisms as well as outcomes on putative disease analogs in rodents and canines, the latter including randomized controlled trials of KD.

With respect to mechanisms, the pathophysiology of the mental disorders covered in this review is not clearly understood, though impaired metabolism due to mitochondrial dysfunction has been identified as an important substrate (34). This is congruent with findings in neurological conditions, Stafstrom and Rho concluding that energy metabolism changes induced by KD in neurological conditions suggest a final common pathway implicating mitochondrial function (26). KD may also influence neuronal plasticity by modifying neural circuits and cellular properties to normalize function (26). Mitochondrial dysfunction may be relevant in some mental disorders including schizophrenia, ASD, and ADHD, whereas the improvements seen in anxiety, depression, and bipolar disorder may be related to alterations of neurotransmitters.

One other possible mediator of the beneficial effects of KD in mental disorders is the effect on sleep. In a study of 18 children with treatment-resistant epilepsy, after 3 months of KD sleep was reported to be enhanced with a pattern of significant reduction in total night sleep, preservation of slow-wave sleep, increased rapid eye movement (REM) sleep, and decrease in sleep stage 2 (79). The mechanisms by which KD affects sleep is unclear (80), and more studies are necessary to confirm reports that certain dietary patterns and foods improve sleep (81).

Sleep problems and mental disorders are codependent conditions that exacerbate each other and lead to impaired quality of life and increased disability (82). Impairments of sleep are a widespread feature of mental disorders. Anxious patients have been found to have significantly less sleep period time, total sleep time, percentage stage REM and percent stage 4 sleep, shorter latency to stage REM, and greater percent stage 1 sleep than healthy controls (83). REM sleep abnormalities including shortening of REM latency, lengthening of the duration of the first REM period, and heightening of REM density are found in patients with depression (84). In patients with inter-episode bipolar disorder, shorter sleep onset latency and increased REM density has been observed (85). A decrease of REM sleep latency in schizophrenia has been described (86). Individuals with ASD have prolonged sleep latency, more frequent nocturnal awakenings, lower sleep efficiency, increased duration of NREM stage 1 sleep, and decreased deeper stages of NREM sleep (87). In ADHD, disturbed sleep architecture has been described including shorter REM latencies, reduced REM sleep, and increased delta sleep percentage (88). It should also be noted that sleep deprivation can precipitate mania in bipolar disorder and seizures in epilepsy (89) and can be used as a treatment for depression (90). The specific effects of KD on these mental disorder-related sleep symptoms has not been studied in detail, but interactions are likely and may be possible mediators of a therapeutic effect.

In epilepsy, KD acts differently to antiepileptic drugs (AED) in seizure prevention. While AED act directly on ion channels and synaptic processes, KD acts through intermediary metabolic pathways (91). Chang et al. showed that an MCT (palm oil and coconut oil) diet, a variation of KD, reduces seizures in children via inhibition on AMPA receptors (12, 92, 93). The questions posed by the literature indicate that the mechanism of action is still unknown, and there may be many potential pathways involved. The mechanism of action appears different from AED and therefore probably psychiatric drugs also, which opens potential avenues for treatment in a manner that may supplement conventional pharmacological treatment approaches. The exact mechanism of action of KD is unclear, and for detailed discussion, see Rogawski et al. (91). Thus, present knowledge indicates that KD exerts its effects on seizure control by mechanisms different from conventional AED and therefore, in psychiatry, this may also be the case although as yet unproven.

There are a number of reasons why the effectiveness of KD in mental disorders remains unproven. In addition to the low number of human studies, the quality of the studies has some significant limitations. Sample sizes are small, there is no control for placebo effects, and the establishment of ketosis is generally lacking with no confirmatory measurement of ketones in three human studies. There are also significant limitations associated with the diet itself including the detailed regimen, unpalatable food choices, side effects, and duration of diet required. There are also no enforced standards as to what constitutes KD in humans with variable lipid:non-lipid ratios reported. KD monotherapy is used in animal models of mental disorders but remains unexamined in human studies. Ten adult patients with epilepsy followed KD monotherapy, and it was concluded that it may be feasible, well tolerated, and an effective long-term alternative (94).

To comply with KD, patients who may be acutely unwell are required to measure food portions to ensure that the macronutrient targets associated with the diet are met, and they may find it difficult to adhere to such a demanding diet (47). This is particularly so for patients with mental disorders where symptoms such as impulsivity in mania, apathy, and reduced appetite in depression, food cravings, and binge eating associated with antipsychotic medications may variously interfere with compliance with KD (95). A mitigating factor to the outcomes in children with epilepsy may be that the diet is typically administered in a hospital setting initially and subsequently, by caregivers.

El-Mallakh and Paskitti have outlined the adverse consequences of KD including constipation, menstrual irregularities, elevated serum cholesterol and triglycerides, hypoproteinemia, hemolytic anemia, elevated liver enzymes, and gall stones (96). Kidney stones have been noted to occur in 1 of 20 children on the diet (97). In a period of almost 2 years, prospective monitoring of 52 children with pediatric epilepsy was conducted. Ten percent of children experienced serious adverse events associated with the diet 1 month after initiation (98). This included presacral and periorbital edema, developmental impairment, and unwanted weight loss in an infant, renal tubular acidosis, viral gastroenteritis, abnormal liver function, and thrombocytopenia. It should be noted that all patients were being treated with concomitant VPA. It was reported in a retrospective study of 158 children with intractable epilepsy that, in 80% emesis, food refusal and hypoglycemia occurred (99).

By definition, KD is confirmed by the production of ketones measured in the blood or urine. In the reviewed literature covering KD in mental disorders, four studies did not report ketone levels, which severely limit comparability across studies and the ability to invoke any consistent mechanism. One study compared whether measuring serum beta-hydroxybutyrate or urinary ketones was superior to monitor KD (100). In humans, it was found that beta-hydroxybutyrate correlated more strongly with a reduction in seizures than urinary ketones; therefore, future studies should measure ketones in the blood. Another issue is that the lipid:non-lipid ratios used were different (see Tables 1 and 2). In a study that compared the efficacy and tolerability of the 3:1 versus the 4:1 lipid:non-lipid ratios, the latter was shown to have a higher seizure-free outcome (2).

One issue when interpreting the results is the levels of evidence in the evidence-based hierarchy. Animal models of mental disorders are considered valuable preclinical tools to investigate the neurobiological basis of a disorder (62). While this may be true, they are nonetheless subject to a number of limitations. One such limitation is the issue of validity, and their use is based on the assumption that humans and animals share basic neurobiological mechanisms associated with the complex behaviors that mimic mental disorders in animals (101).

Another difficulty posed to practitioners is that there are currently no international protocols guiding the administration of the diet; this is something that may be established from future research into KD. There was only one case study that detailed what the participant, diagnosed with schizophrenia, ate, and it was not established whether this individual was in ketosis. In the various studies in humans, outcomes were assessed following dietary durations that varied from 7 days to 2 years.

Further research into the neural correlates of KD is needed to help explain the mechanisms by which it acts. Some suggestions regarding methodologies, provided by Fusar-Poli are elaborated below. Changes in glucose metabolism seen in KD could be examined using positron emission tomography fluorodeoxyglucose. To observe the neural correlates of KD, a combination of electrophysiological measures including EEG and magnetoencephalogram and fMRI/PET to combine the high temporal resolution of the former with the high spatial resolution of the latter may be used (102).

In the neurological literature, a single study, in Alzheimer’s disease, used a synthesized ketogenic compound AC-1202 rather than a KD. AC-1202 is an MCT composed of glycerine and caprylic acid (23). It is not yet clear what role ketogenic pharmacotherapy options might play alongside or as a substitute for KD.

While these animal studies are placing research into KD on a firm footing and identifying some promising leads, on balance the evidence in humans is insufficient to form an opinion as to the efficacy or lack thereof of this intervention in the mental disorders reported. Further basic research to clarify the specifics of dietary manipulation or supplementation required to produce optimum ketosis in specific models is an obvious intermediate step toward studying the effectiveness of the diet in human mental disorders using conventional phases of research including open-label studies and randomized controlled trials.

EB derived the concept of the article from which she received supervision and expert advice in the area of psychiatry from KK and neurology from BT.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.