Mood and anxiety disorders are a major global health burden with a significant unmet need.Reference Forouzanfar, Alexander, Anderson, Bachman, Biryukov and Brauer1 There is an urgent requirement for more effective treatments.Reference Leichsenring, Steinert, Rabung and Ioannidis2 Current treatments include antidepressants, which have variable response ratesReference Cleare, Pariante, Young, Anderson, Christmas and Cowen3 and increasing concerns regarding severe withdrawal experiences,Reference Hengartner, Schulthess, Sorensen and Framer4 and psychological therapies, whose apparent efficacy might be affected by positive publication bias.Reference Driessen, Hollon, Bockting, Cuijpers, Turner and Lu5 In recent years, both preclinical and clinical evidence has emerged that supports the role of diet as an adjunctive therapeutic approach for mood disorders.Reference Marx, Moseley, Berk and Jacka6 Therapeutic options include versions of the Mediterranean diet used in the PREDIMED,Reference Sánchez-Villegas, Martínez-González, Estruch, Salas-Salvadó, Corella and Covas7 SMILESReference Jacka, O'Neil, Opie, Itsiopoulos, Cotton and Mohebbi8 and HELFIMED trialsReference Parletta, Zarnowiecki, Cho, Wilson, Bogomolova and Villani9 and two other Australian trials in young adults.Reference Bayes, Schloss and Sibbritt10,Reference Francis, Stevenson, Chambers, Gupta, Newey and Lim11 Dietary approaches (whether alone or as augmentation) may be especially appealing to people with mood disorders who have mild to moderate symptoms and physical comorbidities, and therefore the potential impact on the wider population and economic factors may be significant.
One proposed dietary approach is carbohydrate reduction (typically with moderate increases in fat and protein), in the form of low carbohydrate (<130 g carbohydrate/day) and very low carbohydrate (<25–50 g/day) ketogenic diets. These have been defined and operationalised (Table 1). Low carbohydrate and ketogenic diets (LC/KD) dating back to the 1860sReference Banting12 and the 1970s Atkins dietReference Atkins13 have shown efficacy in type 2 diabetes, obesity and metabolic syndrome. There is increasing interest, albeit with debate, especially regarding longer-term efficacy and potential adverse effects (such as hypercholesterolaemia, nutritional deficiencies and renal stonesReference Andrewski, Cheng and Vanderpool14,Reference Kang, Chung, Kim and Kim15 ). Interestingly, improvements in mood, energy, sleep, mental clarity and affect stability have been reported with use of LC/KD in type 2 diabetes and obesity.Reference Cucuzzella, Tondt, Dockter, Saslow and Wood16
Source: modified from Accurso et al.Reference Accurso, Bernstein, Dahlqvist, Draznin, Feinman and Fine32
The ketogenic diet has shown some efficacy for medication-resistant paediatric epilepsy, first suggested in 1921.Reference Wilder19 An updated Cochrane reviewReference Martin-McGill, Bresnahan, Levy and Cooper17 reported promising results for this intervention, with greater seizure reduction in those on a ketogenic diet compared with usual care: seizure freedom in up to 55% of children and seizure reduction in up to 85% after 3 months. However, conclusions were limited by the small number of studies and sample sizes and low to very low overall quality of evidence. Nevertheless, ketogenic diets are increasingly used for medication-resistant epilepsy, and there has been an increase in specialist ketogenic diet services.Reference Whiteley, Martin-McGill, Carroll, Taylor and Schoeler24 Together, research from epilepsy as well as studies describing improvements in mood and cognition in type 2 diabetes and obesity has stimulated interest in ketogenic diets for various psychiatric and neurological disorders, including cerebral glioma, Parkinson's disease, multiple sclerosis, chronic fatigueReference Paoli, Rubini, Volek and Grimaldi25,Reference Bostock, Kirkby and Taylor26 and mood disorders.Reference Bostock, Kirkby and Taylor26–Reference Brietzke, Mansur, Subramaniapillai, Banlanzá-Martínez, Vinberg and González-Pinto28
Poor nutrition is a substantial driver of the global increase in non-communicable disorders.Reference Forouzanfar, Alexander, Anderson, Bachman, Biryukov and Brauer1,Reference Molendijk, Molero, Ortuño Sánchez-Pedreño, Van der Does and Angel Martínez-González29–Reference Lane, Davis, Beattie, Gómez-Donoso, Loughman and O'Neil31 The pro-inflammatory and high carbohydrate content (typically >300 g/dayReference Accurso, Bernstein, Dahlqvist, Draznin, Feinman and Fine32) of Western-pattern diets, especially high refined carbohydrates and ultra-processed foods, may be key factors.Reference Lane, Davis, Beattie, Gómez-Donoso, Loughman and O'Neil31
Importantly, it should be noted that the term ‘carbohydrates’ is now colloquially used to refer to highly refined carbohydrates, which make up a substantial part of Western diets. However, foods that are known to impart substantial health benefits due to their fibre contentReference Reynolds, Mann, Cummings, Winter, Mete and te Morenga33 as well as dietary components used by gut bacteria to form molecules essential to health, such as polyphenols and resistant starch, including starchy vegetables, different whole-grain cereals and legumes, are also high in carbohydrates. These foods form a substantial part of Mediterranean diets, which have been linked to positive health outcomes.Reference Martinez-Lacoba, Pardo-Garcia, Amo-Saus and Escribano-Sotos34 References to ‘carbohydrates’ throughout this review should be read with this important distinction in mind.
LC/KD are plausible therapies for mood disorders for several reasons. The relationship between mood disorders and carbohydrate intake is complex, with both acute and habitual effects at play. At a population level, sugar intake correlates with depression rates.Reference Westover and Marangell35 High glycaemic index (GI) diets appear to negatively affect mood.Reference Christensen36–Reference Knüppel, Shipley, Llewellyn and Brunner45 Furthermore, ‘comfort eating’ is common in mood disorders and carbohydrate cravings are often present in seasonal affective disorderReference Jacobsen, Wehr, Sack, James and Rosenthal46 and atypical depression, which is common in bipolar disorder.Reference Rahe, Baune, Unrath, Arolt, Wellmann and Wersching47 Reducing carbohydrate intake might therefore alleviate mood disorder symptoms in some individuals. Indeed, animal models of depression suggest that a ketogenic diet might exert an antidepressant or anxiolytic effect.Reference Murphy, Likhodii, Nylen and Burnham48 Case studies have also reported amelioration of psychotic symptoms following initiation of a ketogenic diet.Reference Sarnyai and Palmer49 However, randomised controlled trials (RCTs) are required to establish efficacy in reducing depression, anxiety and psychotic symptoms in psychiatric populations.
Although the efficacy of LC/KD for mood and anxiety disorders has not yet been established, public interest is increasing and clinicians may already encounter individuals utilising such approaches.Reference Campbell and Campbell50 A self-report uncontrolled survey of low carbohydrate diets in 1580 patients with a variety of conditions, including obesity and type 2 diabetes, reported significant improvements in mood, anxiety and energy and a reduction in antidepressant use (although psychiatric diagnoses were not provided).Reference Cucuzzella, Tondt, Dockter, Saslow and Wood16 In contrast, a systematic review of LC/KD RCTs in obese and overweight adults without epilepsy or mood disorders (eight studies, n = 532, duration 8 weeks to 1 year) reported no overall psychological benefits.Reference El Ghoch, Calugi and Dalle Grave51 Two RCTs have also examined mood and behavioural effects of a ketogenic diet in paediatric epilepsy: Ijff et al reported reductions in anxious and mood-disturbed behaviour in those on a ketogenic diet,Reference IJff, Postulart, Lambrechts, Majoie, de Kinderen and Hendriksen52 whereas Lambrechts et al found a tendency towards an increase in mood problems.Reference Lambrechts, Bovens, De la Parra, Hendriksen, Aldenkamp and Majoie53
Given this background, the aim of the current study was to systematically review low carbohydrate and ketogenic interventions in adults with mood and anxiety disorders to direct further avenues for research and to highlight uncertainties for clinical practice.
Method
This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelinesReference Moher, Liberati, Tetzlaff, Altman, Altman and Antes54 and registered with the International Prospective Register of Systematic Reviews (PROSPERO: CRD42019116367). The following adjustments to the methods set out in the protocol were made during the review: (a) English language articles only were considered; (b) studies did not need to include a non-exposed comparison group to be eligible for inclusion, to widen the scope and generalisability of the review; (c) an adapted version of the Newcastle–Ottawa Scale was used for the risk of bias assessment,Reference Wells, Shea, O'Connell, Peterson, Welch and Losos55 as it was more appropriate for the included study designs; and (d) grey literature was excluded in order to improve the quality of included studies.
Eligibility criteria
Eligibility criteria: (a) primary research article; (b) published between database inception to 6 September 2022; (c) published in a peer-reviewed journal; (d) human adult participants (>18 years of age); (e) published in the English language; (f) individuals with mood or anxiety disorders, defined here as depression, anxiety, bipolar disorder and schizoaffective disorder; (g) low carbohydrate diet or ketogenic diet as the intervention. Exclusion criteria were papers in which the population had a primary neurodevelopmental disorder diagnosis (such as autism spectrum disorder or attention-deficit hyperactivity disorder) and animal studies.
Information sources and search strategy
The search strategy was developed based on the research question ‘Do low carbohydrate or ketogenic diets confer mood benefits in mood and anxiety disorders?’ and the following PICOS (participants, interventions, comparisons, outcomes, study design) criteria:
(a) participants: adults with depression, anxiety, bipolar disorder or schizoaffective disorder
(b) interventions: low carbohydrate or ketogenic diets
(c) comparisons: if reported, ‘normal’ or ‘moderate’ carbohydrate diet, or Western-pattern diet, or placebo, or usual care, or befriending, or psychological support
(d) outcomes: reports of response, remission, relapse of symptoms (where reported, as defined by validated measures); adverse events; attrition
(e) study design: RCTs; if RCTs were not available, cohort studies, case series, case reports.
A systematic search of MEDLINE, Embase, PsycINFO and Cochrane was performed on 6 September 2022. No lower date search limit was set. The search strategy is detailed in Supplementary file 1, available at https://dx.doi.org/10.1192/bjo.2023.36.
Study selection
Two reviewers (D.M.D. and M.H. and subsequently D.M.D. and J.K.-G.) independently assessed all retrieved records for inclusion using Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia; www.covidence.org). First, reviewers screened titles and abstracts for eligibility. Any potentially eligible or ambiguous records were retained for the second round of screening, in which full texts were examined. Any discrepancies in the final articles to be included were discussed in a consensus meeting.
Data extraction
Two authors (of D.M.D., J.K.-G., M.H. and W.M.) extracted the following information from included studies using Covidence software: author/date, study design, sample size, population characteristics (age, gender, comorbidities), type and duration of dietary intervention, adverse events and mental health-related outcomes, including measures of depression, anxiety and mood, and psychosis rating scales. The primary outcome was a clinically significant change in symptoms and/or change from baseline to last observation for relevant rating scales.
Study risk of bias assessment
An adapted version of the Newcastle–Ottawa Scale was used for the risk of bias assessment,Reference Wells, Shea, O'Connell, Peterson, Welch and Losos55 using items appropriate for cross-sectional, observational, cohort and non-randomised interventional study designs.
Results
Characteristics of included studies
The results of the search strategy are summarised in Fig. 1. The search yielded 1377 articles, of which 48 full-text articles were assessed for eligibility. Of these studies, 36 were excluded. The 12 eligible studies are summarised in Table 2, stratified by study design. There were no RCTs. The studies comprised nine case reports, two cohort studies and one observational study. Bipolar disorder was the most studied mood disorder (n = 6 studies). All studies stated that the intervention was a ketogenic diet. However, there is no widely accepted definition of a ketogenic diet, carbohydrate intake was not reported in five studies and confirmation of ketosis was incomplete (blood ketone testing in two studies, urinary ketone testing in five, no confirmation in five). It is therefore possible that some of the interventions may have been a low carbohydrate rather than a ketogenic diet.
β-HB, beta-hydroxybutyrate; BDI, Beck Depression Inventory; GAD-7, seven-item Generalised Anxiety Disorder questionnaire; HRSD, Hamilton Rating Scale for Depression; MADRS, Montgomery–Åsberg Depression Rating Scale; MDD, major depressive disorder; n.r., not reported; PANSS, Positive and Negative Syndrome Scale; PHQ-9, nine-item Patient Health Questionnaire (depression module).
Risk of bias assessment
Cohort and observational studies were assessed for risk of bias (Supplementary file 2). Owing to the heterogeneity in study designs, we did not provide a quality score for each study and instead provided a qualitative assessment.Reference Page, Mckenzie and Higgins56 None of the studies employed representative sampling techniques and confounders were not controlled for. Changes in anxiety, depression and/or psychotic symptoms were measured via validated questionnaires in four studies.Reference Shegelman, Carson, McDonald, Henry-Barron, Diaz-Arias and Cervenka57–Reference Danan, Westman, Saslow and Ede59 One study additionally measured depression symptoms and illness severity via clinical interview.Reference Danan, Westman, Saslow and Ede59
Synthesis of results
Considering the methodological heterogeneity between studies, a quantitative appraisal was not possible. Findings were therefore synthesised narratively, clustered by study type and then stratified as either ‘no efficacy’ or ‘possible efficacy’ of the intervention. Discontinuation effects and adverse events were described where reported (for clarity, these have been grouped together, given the limited data).
Case reports
No efficacy of LC/KD intervention
Two case reports reported no apparent benefits of the intervention (duration 4 weeks each). One report described a participant with panic disorder (n = 1)Reference Ehrenreich60 on a ketogenic diet who experienced negative effects (internal ‘shakiness’, increase in anxiety and recurrence of panic attacks, despite increasing sertraline dose). These symptoms resolved on discontinuation of the ketogenic diet. The other case report described an unsuccessful attempt at a ketogenic diet in a hospital in-patient with rapid cycling bipolar disorder (mania predominant) (n = 1).Reference Yaroslavsky, Stahl and Belmaker61 Ketosis was not achieved despite reported adherence to the diet, so this may have been a low carbohydrate diet rather than a ketogenic one.
Possible efficacy of LC/KD intervention
Seven case reports reported possible efficacy (duration of diet between 1 month and 3 years). Three described substantial improvements in symptoms in individuals with bipolar disorder (n = 4), including greater mood stability, reductions in frequency of mood episodes and decreased anxiety.Reference Saraga, Misson and Cattani62–Reference Chmiel64 Four patients discontinued or reduced their antipsychotic or mood stabiliser medication. Ketosis was measured by blood β-hydroxybutyrate in one reportReference Chmiel64 and by urinary ketones in twoReference Saraga, Misson and Cattani62,Reference Phelps, Siemers and El-Mallakh63 reports. The other four case reports described a female with type 2 diabetes and comorbid major depressive disorder,Reference Cox, Gibas, Salisbury, Gomer and Gibas65 a female with schizophrenia and comorbid depression,Reference Kraft and Westman66 a female with a mood disorder and comorbid emotion dysregulation, body dysmorphic disorder and an eating disorder,Reference Pieklik, Pawlaczyk, Rog and Karakuła-Juchnowicz67 and two people (one male, one female) with schizoaffective disorder and comorbid major depressive disorder.Reference Palmer58 Reported benefits included improved mood, energy, concentration and cognition, as well as reduced psychotic symptoms in those with schizoaffective disorder.
Cohort studies
No efficacy of LC/KD intervention
No studies.
Possible efficacy of LC/KD intervention
Possible efficacy was reported in a cohort study of ketogenic and low carbohydrate diets in out-patients with anxiety, depression and ‘dysperception’ (n = 73; duration not reported)Reference Kunin68 and in a study of a ketogenic diet in in-patients with bipolar II disorder, major depressive disorder or schizoaffective disorder (n = 31, mean duration 59.1 days).Reference Danan, Westman, Saslow and Ede59 The first study reported improvements in anxiety, depression and ‘dysperception’ in 82% of participants. Similarly, the second study reported improvements in self- and clinician-rated depression, as well as overall severity of illness in participants with mood disorders; further, around two-thirds of participants reduced the number or dose of psychotropic medications by the end of the intervention.
Observational studies
No efficacy of LC/KD intervention
No studies.
Possible efficacy of LC/KD intervention
In the only observational study in the review, Campbell & CampbellReference Campbell and Campbell50 reported an analytic study of posts on online forums about a ketogenic diet versus omega-3 supplementation or a vegetarian diet from people with bipolar disorder (n = 274; reporting of diet duration incomplete but where available it was between 1 month and >12 months). Remission or significant mood stabilisation was reported in 56.4% of posts discussing a ketogenic diet, compared with 14.9% of posts on a vegetarian diet or omega-3 supplementation. Posts discussing mood destabilisation were few overall, but more commonly associated with a vegetarian diet or omega-3 supplementation than a ketogenic diet.
Effects of discontinuation of LC/KD intervention
Of the ten studies that reported symptom improvement with LC/KD, only four reported information on symptom changes on stopping LC/KD. Campbell & CampbellReference Campbell and Campbell50 reported recurrence of bipolar disorder symptoms in 7.1% of participants on stopping a ketogenic diet and KuninReference Kunin68 reported recurrence of symptoms in 82% of participants with depression, anxiety and ‘dysperception’ on cessation of a low carbohydrate diet. Two case reports reported mixed results. Kraft & WestmanReference Kraft and Westman66 reported no recurrence of psychotic symptoms during dietary relapses in a person with schizophrenia and comorbid depression; however, PalmerReference Palmer58 reported recurrence of psychotic symptoms during dietary relapse in two individuals with schizoaffective disorder, despite continuing antipsychotic medication in one of them. Symptoms resolved when ketosis was induced; however. an increase in the individual's antipsychotic medication dosage may also explain this improvement.
Adverse events
Adverse events were not reported systematically (Table 2). Six studies did not report on adverse events.Reference Palmer58,Reference Yaroslavsky, Stahl and Belmaker61,Reference Saraga, Misson and Cattani62,Reference Chmiel64–Reference Kraft and Westman66 Two studies reported no adverse events (n = 3).Reference Phelps, Siemers and El-Mallakh63,Reference Pieklik, Pawlaczyk, Rog and Karakuła-Juchnowicz67 Four studies reported adverse events, including: transient symptoms associated with keto-adaptation (e.g. fatigue, nausea, headache, palpitations) (n not recorded),Reference Campbell and Campbell50,Reference Danan, Westman, Saslow and Ede59,Reference Kunin68 ‘shakiness’, increase in anxiety and recurrence of panic attacks (n = 1),Reference Ehrenreich60 fat intolerance (n = 2) and gastrointestinal symptoms (n = 2).Reference Danan, Westman, Saslow and Ede59
Discussion
This systematic review examined the efficacy of low carbohydrate and ketogenic diets (LC/KD) in individuals with mood and anxiety disorders. Despite anecdotal reports and biological plausibility, little research has been conducted to date and no high-grade evidence was found. Heterogeneity and data quality limit interpretation of the included studies, and statistical analysis including meta-analysis was not possible, leading to a narrative review.
It should be noted that nutritional intake and measurements of ketosis were variable and incomplete, and although the interventions were stated to be ketogenic diets, there was no consistent threshold for ketone levels to define ketosis and it is possible that some individuals were on low carbohydrate rather than ketogenic diets, either throughout the duration of the study or periodically, and did not reach the required threshold for ketosis.
Although two studies suggested no benefits of LC/KD, several case reports, cohort studies and the observational study suggest possible efficacy of a ketogenic diet in bipolar disorder, schizoaffective disorder, anxiety disorders and unipolar depression. Reported therapeutic effects of a ketogenic diet include mood stability, increased energy and concentration, and reductions in symptoms of anxiety, depression and psychosis. Relapse occurred in several individuals on discontinuation of the diet. Adverse effects were not systematically reported but included fatigue, nausea and headaches on induction of ketosis, although this was not universal.
Potential benefits of a ketogenic diet reported here concur with previous research exploring self-reported benefits of low carbohydrate diets in other populations.Reference Cucuzzella, Tondt, Dockter, Saslow and Wood16,Reference Ernst and Shelley-Tremblay69 The present paper expands on reviews by Bostock et alReference Bostock, Kirkby and Taylor26 and Brietzke et alReference Brietzke, Mansur, Subramaniapillai, Banlanzá-Martínez, Vinberg and González-Pinto28 by incorporating low carbohydrate interventions (not just a ketogenic diet) and a wider clinical context: this may increase generalisability to community samples of individuals with mood disorders. In contrast, in individuals without mood disorders, El Ghoch et alReference El Ghoch, Calugi and Dalle Grave51 found no overall evidence of psychological benefits of LC/KD (eight studies, n = 532; duration 8 weeks to 1 year). However, in El Ghoch et al's review: (a) two studies did not meet criteria for a low carbohydrate diet,Reference Dalle Grave, Calugi, Gavasso, El Ghoch and Marchesini70,Reference Galletly, Moran, Noakes, Clifton, Tomlinson and Norman71 (b) ketosis was measured in only five studiesReference Rosen, Hunt, Sims and Bogardus72–Reference Saslow, Kim, Daubenmier, Moskowitz, Phinney and Goldman76 and (c) several included studies utilised the 24 h Food Frequency Questionnaire, the validity of which has been questioned.Reference Archer, Pavela and Lavie77 Alternatively, it might be that LC/KD are efficacious only in individuals with clinical mood disorders and thus more severe psychopathology, but not in non-clinical samples with milder symptoms. It is also possible that LC/KD are more useful in those with greater metabolic burden, such as threshold metabolic syndrome.Reference Volek and Feinman78
Context and mechanisms
Any response to LC/KD may be modulated via multiple mechanistic pathways. For example, LC/KD have been suggested to reduce inflammation, which is increasingly acknowledged to be involved in the development of depression and, in particular, treatment-resistant depression.Reference Firth, Marx, Dash, Carney, Teasdale and Solmi79–Reference Nettis, Lombardo, Hastings, Zajkowska, Mariani and Nikkheslat82 A systematic reviewReference Buyken, Goletzke, Joslowski, Felbick, Cheng and Herder83 showed that low glycaemic index and low glycaemic load diets were associated with reduced inflammatory markers in 5 out of 9 observational studies; 3 of 13 interventional trials showed significant anti-inflammatory effects and 4 suggested beneficial trends. Reduced inflammation correlates with reduced depression in obesity and overweight,Reference Perez-Cornago, de la Iglesia, Lopez-Legarrea, Abete, Navas-Carretero and Lacunza84,Reference Baumeister, Russell, Pariante and Mondelli85 which may be related to reduced inflammatory adipokines.Reference Mancuso86,Reference Forsythe, Phinney, Fernandez, Quann, Wood and Bibus87
Mitochondrial energy generation – which can be affected by ketosis – is also important, because mood disorders, especially bipolar disorder, are associated with abnormal mitochondrial energy generation.Reference Morris, Puri, Carvalho, Maes, Berk and Ruusunen88,Reference Campbell and Campbell89 Depression is characterised by reduced energy generation, and mania is characterised by increased mitochondrial biogenesis.Reference Morris, Walder, McGee, Dean, Tye and Maes90 A ketogenic diet has the potential to increase the efficiency of mitochondrial biogenesis and may target this core pathway.Reference Morris, Puri, Carvalho, Maes, Berk and Ruusunen88,Reference Morris, Maes, Berk, Carvalho and Puri91,Reference Morris, Walder, Berk, Marx, Walker and Maes92
Furthermore, improvements in insulin signalling conferred by LC/KD (as evident in treatment of obesity and type 2 diabetes) may be associated with mood benefits. Calkin et al showed that individuals with bipolar disorder and comorbid type 2 diabetes or insulin resistance had three times higher odds of chronicity and rapid cycling and more than eight times the odds of lithium resistance, even after controlling for antipsychotic use and body mass index (BMI).Reference Calkin, Ruzickova, Uher, Hajek, Slaney and Garnham93 Subsequently, in a first RCT, Calkin et al showed that reversing insulin resistance by adding metformin in treatment-resistant bipolar depression had a large effect size on clinician-rated depressive symptoms – but only in participants whose insulin resistance was reversed.Reference Calkin, Chengappa, Cairns, Cookey, Gannon and Alda94 Reversing insulin resistance is clearly a plausible outcome of a low carbohydrate diet, not requiring ketogenic diet levels of carbohydrate reduction. In patients at risk, low carbohydrate diets can be advocated for prevention of obesity, metabolic syndrome and type 2 diabetes, with the possible additional benefit of an antidepressant effect in bipolar disorders. Thus, the role of insulin resistance in mood disorders may justify the application of LC/KD interventions in this population.
Ketosis has neuroprotective mechanisms in epilepsy,Reference Masino and Rho95 so it is possible that ketosis may also confer positive mood outcomes in non-epileptic disorders. As proof of concept, short-term fasting (generating ketosis) may improve mood and induce mild euphoria in individuals with mood disorders.Reference Fond, Macgregor, Leboyer and Michalsen96 Longer-term ketosis-related mood stabilisation in bipolar disorder is reflected in the self-reports analysed by Campbell & Campbell.Reference Campbell and Campbell50 Putative mechanisms of ketosis on mood may involve several factors, including increased fatty acid synthesis and oxidation, neurotransmitters, ion channels, mitochondrial genesis, cell signalling, second messengers and reduced oxidative stress.Reference Newman and Verdin97,Reference Shimazu, Hirschey, Newman, He, Shirakawa and Le Moan98 Notably, the seizure-reducing effects of ketosis in epilepsy have not been clarified. Ketosis also induces brain-derived neurotrophic factor (BDNF) gene expression;Reference Sleiman, Henry, Al-Haddad, El Hayek, Abou Haidar and Stringer99 this is important as a low BDNF level adversely affects neuroplasticityReference Autry and Monteggia100 and BDNF increases with effective treatment.Reference Post101 Weight loss associated with a ketogenic diet might also be a relevant mechanism, although reports of significant mood improvement as early as 4 days into a ketogenic diet, which precedes any significant weight loss, is consistent with ketosis-mediated effects on mood rather than weight loss per se. A further factor potentially playing a role in mediating effects of a ketogenic diet on mood is reduced appetite,Reference Sumithran, Prendergast, Delbridge, Purcell, Shulkes and Kriketos102 as discussed in more detail below.
Limitations
There are several limitations to this review, partly related to the low number of studies on this topic and their general poor quality (for example, lack of non-exposed control groups and consideration of confounding factors). The broad inclusion criteria, including lower levels of evidence, may have resulted in heterogeneity between studies and variable data quality.
We pragmatically grouped depression, anxiety, bipolar disorders and schizoaffective disorders together, given some overlapping symptomatology, because of the paucity of LC/KD evidence for individual disorders. A case could also be made against the inclusion of studies with limited information on psychiatric diagnoses,Reference Shegelman, Carson, McDonald, Henry-Barron, Diaz-Arias and Cervenka57,Reference Kunin68 as the results of these studies may not be generalisable to individuals with mood disorders without such comorbidities.Reference Paoli, Bosco, Camporesi and Mangar103
The effects of LC/KD on mood disorders might also involve confounding factors, which were rarely assessed in the studies included in this review. These include concurrent pharmacotherapy, reduced alcohol intake (high carbohydrate content), reduced hunger (ketosis effect on appetite hormones),Reference Paoli, Bosco, Camporesi and Mangar103 weight loss, improved sleep and (in those on a well-formulated low carbohydrate diet) presumed increased intake of folate, micronutrients, vitamin D and omega-3 fatty acids (LC/KD typically increase intake of leafy green vegetables and fish).
The exclusion of certain foods may also be an important confounder. For example, reduced gluten intake in LC/KD has been suggested to be a confounder in a limited number of studies, as there are reports of anti-gluten antibodies in psychosisReference Lachance and McKenzie104 and bipolar disorder.Reference Dickerson, Stallings, Origoni, Vaughan, Khushalani and Alaedini105 Furthermore, an unnecessary gluten-free diet (which LC/KD might entail) might increase the risk of nutritional deficienciesReference Naik, Seidner and Adams106 and gastrointestinal and cardiovascular side-effects.Reference Nash and Slutzky107,Reference Lebwohl, Cao, Zong, Hu, Green and Neugut108
Another potential confounder is the potentially low content of fermentable oligosaccharides, disaccharides, monosaccharides and polyols (FODMAPs) in LC/KD (depending on the dietary formulation). FODMAPs, poorly absorbed short-chain carbohydrates (for example in bread), may cause bloating and fatigue in some individuals and are implicated in the symptomatology of irritable bowel syndrome (IBS): FODMAP reduction or exclusion improves IBS.Reference Gibson, Muir and Newnham109 The relevance here is that IBS and various psychiatric disorders (notably, anxiety disorders and, possibly, current major depression) are often comorbid.Reference Fadgyas-Stanculete, Buga, Popa-Wagner and Dumitrascu110 Interestingly, the omission of FODMAPs confounds gluten-free diet trials in IBSReference Gibson, Muir and Newnham109 and although there is uncertainty regarding the psychiatric effects of FODMAPs,Reference Aranburu, Matias, Simón, Larretxi, Martínez and Bustamante111 it could be argued that omitting FODMAPs might confound any apparent psychiatric effects of LC/KD.
Three of the included studiesReference Palmer58,Reference Yaroslavsky, Stahl and Belmaker61,Reference Chmiel64 also report intake of exogenous ketogenic food supplements: medium-chain triglycerides. The role of exogenous ketogenic food supplements is unclearReference Harvey, Schofield and Williden112 and it must be noted that these compounds are high in saturated fatty acids, which may increase the risk of cardiovascular disease.Reference Maki, Dicklin and Kirkpatrick113
Finally, no studies assessed gut microbiota, which may affect mood and/or anxiety.Reference Maki, Dicklin and Kirkpatrick113 A ketogenic diet may be associated with changes in alpha diversity and beneficial microbial metabolites such as short-chain fatty acids,Reference Attaye, van Oppenraaij, Warmbrunn and Nieuwdorp114 although the data are limited and clinical implications uncertain. Further discussion is beyond the scope of this review.
Implications for clinical practice
Clinicians in primary care, mental health services and dietetics/nutrition may increasingly encounter patients who wish to follow LC/KD for depression, anxiety, bipolar disorder or schizoaffective disorder. Shared decision-making and good communication of the current absence of data able to clearly support or refute the efficacy of a ketogenic diet is important. Some of these issues are discussed below and in Table 3.
Low carbohydrate diets – and especially ketogenic diets – are by nature highly restricted diets, whereas ‘whole of diet’ approaches are preferable and more likely to be sustainable in people with mood disorders.Reference Opie, Itsiopoulos, Parletta, Sanchez-Villegas, Akbaraly and Ruusunen115 As illustrated in Fig. 2, adherence to these diets, especially ketogenic diets, may be challenging, leading to either premature cessation (before potential benefits) or after therapeutic nutritional ketosis has been reached (which could equally apply to other treatment modalities).
Adverse effects include (a) early ‘flu-like’ symptoms during induction of ketosis,Reference Harvey, Schofield and Williden112,Reference Bostock, Kirkby, Taylor and Hawrelak116 predominantly due to early electrolyte changes (especially sodium diuresis), contributing to fatigue, constipation, cramps, palpitations and postural hypotension, and (b) perhaps renal stones, gout, osteoporosis and cardiovascular and gastrointestinal disease (see below) in the longer term.
The data on longer-term LC/KD adherence in various clinical populations remain uncertain. In epilepsy, a Cochrane review (n = 932 children and adults) found that adults following a ketogenic diet may be up to five times more likely to drop out of studies compared with those receiving usual care.Reference Martin-McGill, Bresnahan, Levy and Cooper17 In contrast, an online nutritional ketosis programme for adults with type 2 diabetes (n = 349) reported 74% adherence at 2 years (n = 194/262), compared with 78% adherence for those in usual care (n = 68/87).Reference Athinarayanan, Adams, Hallberg, McKenzie, Bhanpuri and Campbell117 The level of support provided to help participants follow the dietary intervention is likely to affect attrition rates, but this factor has not been systematically studied in relation to ketogenic diets. Further, these data have uncertain generalisability to patients with mood disorders, as it is possible that LC/KD adherence may be more challenging for individuals with such disorders. Pertinently, carbohydrate cravings and impulsivity present in many psychiatric disordersReference Kulacaoglu and Kose118 would presumably increase the chance of consumption of high carbohydrate foods, with resultant dietary relapse (as shown in Fig. 2). Furthermore, repeated attempts at LC/KD with initial efficacy but relapse after challenges might exacerbate feelings of guilt and failure. It would therefore be crucial to develop strategies to overcome these challenges in clinical populations.
Several practical aspects, especially of ketogenic diets, warrant consideration. First, daily carbohydrate counting (which may be required to help maintain ketosis) might exacerbate disordered eating and/or obsessive traits. Second, although confirmation of ketosis might be preferable, especially for individuals new to ketogenic diets, this might unnecessarily ‘medicalise’ normal eating patterns.
In addition, measurements of ketosis are temporally variable, may be imprecise (especially in mild nutritional ketosis, in contrast to diabetic ketoacidosis) and may pose logistical challenges. Methods for confirming ketosis include urine (acetoacetate), blood (β-hydroxybutyrate) and breath (acetone) analyses. Urinary ketone testing strips, although popular (they are low cost, non-invasive and easy to use), are unreliable: in mild dietary ketosis sensitivity is 35–76% and specificity 78–100%.Reference Gibson, Eroglu, Rooney, Harper, McClintock and Franklin119 Blood ketone monitoring (venous, capillary) is more reliable: depending on the magnitude of ketosis, sensitivity is 98–100% and specificity 85–93.3%;Reference Brooke, Stiell and Ojo120 however, it may be impractical in the real world because of the cost and discomfort of blood tests or finger-prick testing (for related reasons, in diabetes, regular finger-prick glucose testing is being replaced by indwelling interstitial glucose devices,Reference Bailey, Bode, Christiansen, Klaff and Alva121 but interstitial ketone monitoring devices are not currently available). Breath ketone analysis is a relatively recent technology with uncertain accuracy and some devices are not registered with the Food and Drug Administration.Reference Alkedeh and Priefer122
The involvement of dietitians was an important part of clinical trials of dietary interventions in participants with depression such as the SMILES and HELFIMED trials.Reference Jacka, O'Neil, Opie, Itsiopoulos, Cotton and Mohebbi8,Reference Parletta, Zarnowiecki, Cho, Wilson, Bogomolova and Villani9 This is also supported by a recent meta-analysis that reported that dietitian-led intervention trials showed a greater improvement in depressive symptoms.Reference Firth, Marx, Dash, Carney, Teasdale and Solmi79 This may be resource-intensive in naturalistic settings, but group sessions or internet-based delivery may be acceptable alternatives.Reference Saslow, Mason, Kim, Goldman, Ploutz-Snyder and Bayandorian123
Importantly, a paucity of longitudinal studies has led to uncertainty and debate regarding aspects of long-term LC/KD safety, especially potential cardiovascular disease (CVD) and gastrointestinal risks from increased saturated fatty acid intake and reduced dietary fibre. Increased saturated fatty acid intake has an unpredictable effect on lipids (a surrogate marker for CVD).Reference Naude, Schoonees, Senekal, Young, Garner and Volmink124–Reference Norwitz, Mindrum, Giral, Kontush, Soto-Mota and Wood128 Reduced dietary fibre has potentially negative effects on gut metabolites and microbiota, which may increase the risk of CVD, gastrointestinal and other diseases.Reference Russell, Gratz, Duncan, Holtrop, Ince and Scobbie129–Reference Aune, Keum, Giovannucci, Fadnes, Boffetta and Greenwood131
Finally, the potential effect of LC/KD on medications is important. Careful monitoring is essential and medication amendments may be warranted. For example, a ketogenic diet may affect pharmacokinetics and valproate levels.Reference Spilioti, Pavlou, Gogou, Katsanika, Papadopoulou-Alataki and Grafakou132,Reference Heo, Kim and Chang133 Ketogenic diet-related diuresis (especially during keto-adaptation) may cause electrolyte changes that could affect lithium levels. These factors might pose additional challenges if individuals have intermittent adherence to LC/KD.
Recommendations for further research
Robust RCTs are required to investigate the effects of low carbohydrate and ketogenic dietary interventions (with cross-over) on short- and longer-term outcomes in dysthymia, generalised anxiety disorder, major depressive disorder, atypical depression, bipolar disorder and schizoaffective disorder. Rating scale subgroup analysis might identify specific LC/KD effects on mood, anxiety, energy and sleep. The comparison of low carbohydrate diets with ketogenic diets is especially pertinent because, if efficacious, the less restrictive nature of low carbohydrate diets could increase generalisability. It would also be important to determine whether comorbid obesity/metabolic syndrome affects efficacy, adherence and relapse rates. Continuous interstitial glucose monitoring, such as in the large (n = 1100) multinational PREDICT I nutrition study,Reference Berry, Valdes, Davies, Delahanty, Drew and Chan134 combined with daily self-rated mood measures (for example, the True Colours app135) might enable a fine degree of personalisation. This might, for example, show that individuals with impaired carbohydrate metabolism (a larger post-prandial glucose incremental area under the curveReference Allison, Paultre, Maggio, Mezzitis and Pi-Sunyer136) are more likely to respond to LC/KD, but (perhaps) not those with normal response.
These studies may be informed by the first RCT of a ketogenic diet in psychosis (NCT03873922), an open label trial of a ketogenic diet in euthymic bipolar disorder and schizophrenia (NCT03935854) and a feasibility study of a ketogenic diet in bipolar disorder (ISRCTN61613198). LC/KD effects on pharmacotherapy also require further study.
Neuroimaging may be informative. In people with epilepsy treated with a ketogenic diet, magnetic resonance spectroscopy (MRS) shows an increase in cerebral energy,Reference Seymour, Bluml, Sutherling, Sutherling and Ross137,Reference Pan, Bebin, Chu and Hetherington138 which is thought to be a potential mediator of ketogenic diet efficacy. MRS might therefore also be relevant to ketogenic diets in populations with mood disorders, especially bipolar disorder, given the anomalies of mitochondrial biogenesis.Reference Morris, Puri, Carvalho, Maes, Berk and Ruusunen88,Reference Morris, Maes, Berk, Carvalho and Puri91,Reference Morris, Walder, Berk, Marx, Walker and Maes92 Positron emission tomography or functional magnetic resonance imaging might also inform any ketosis-mediated neuroplasticity and neuroprotection, including changes in higher cortical function and cognition.Reference Fusar-Poli, Cortesi and Veggiotti139 Imaging research could also target the putative anti-inflammatory actions of LC/KD, focusing on changes in microglial activationReference Baumeister, Russell, Pariante and Mondelli85,Reference Valkanova, Ebmeier and Allan140,Reference Mondelli, Vernon, Turkheimer, Dazzan and Pariante141 with translocator protein densityReference Setiawan, Wilson, Mizrahi, Rusjan, Miler and Rajkowska142 in the dorsal anterior cingulate cortex and prefrontal cortex (PFC). Functional BDNF imaging (amygdala, hippocampus, PFC) would add further information on potential mechanisms of LC/KD.Reference Cole, Weinberger, Mattay, Cheng, Toga and Thompson143,Reference Kuhn, Popovic and Pezawas144 Measurement of lactate might be the simplest biomarker of bioenergetics.Reference Kuang, Duong, Jeong, Zachos and Andreazza145
Further research on LC/KD in mood disorders should include appetite hormones (including insulin, leptin and ghrelin), because of the interplay between LC/KD, appetite, insulin sensitivity and mood disorders.Reference Forsythe, Phinney, Fernandez, Quann, Wood and Bibus87,Reference Paz-Filho, Mastronardi, Wong and Licinio146,Reference Zarouna147 These factors might also link with inflammation noted above. Briefly, the role of appetite hormones and mood disorders is complex and may differ between disorders.Reference Zarouna147 For example, Cordas et al showed that reduced leptin is present in depression, especially atypical and bipolar depression; additionally, leptin is associated with current depression.Reference Cordas, Gazal, Schuch, Spessato, Branco and Jansen148 Conversely, Pasco et al showed that leptin is elevated in women with a history of unipolar major depression and elevated leptin predicts future depressive disorder.Reference Pasco, Jacka, Williams, Henry, Nicholson and Kotowicz149
Notwithstanding this, leptin has multiple roles in glucose homeostasis, including improving insulin sensitivity,Reference Paz-Filho, Mastronardi, Wong and Licinio146 which, as noted above, has been shown to be important in treating bipolar depression.Reference Calkin, Ruzickova, Uher, Hajek, Slaney and Garnham93,Reference Calkin, Chengappa, Cairns, Cookey, Gannon and Alda94 LC/KD reduces leptin while also improving sensitivity to leptin.Reference Volek, Phinney, Forsythe, Quann, Wood and Puglisi150 The role of ghrelin in mood disorders is unclear,Reference Zarouna147 but ketosis rapidly suppresses ghrelin, reducing appetite within a few days.Reference Sumithran, Prendergast, Delbridge, Purcell, Shulkes and Kriketos102
Thus, the degree and rate of change of appetite hormones and mood disorder symptoms in various disorders during LC/KD interventions would be instructive. These findings may have important translation potential (Table 3) and might lead to formulation of dietary recommendations for mood disorders (as suggested for those with chronic pain).Reference Rondanelli, Faliva, Miccono, Naso, Nichetti and Riva151
Finally, the effect of LC/KD on overall healthcare costs in mood disorders would also be of interest, as these costs have been shown to be reduced by LC/KD in people with type 2 diabetes.Reference Unwin, Delon, Unwin, Tobin and Taylor152
This review was conducted in the context of debate around many aspects of nutrition, especially ‘low carb’ diets, and mounting scientific and public interest. There is a very limited and low-quality evidence base on the efficacy of KD in mood disorders. Efficacy and adherence are unclear, and there are concerns about potential adverse effects. Nevertheless, some individuals already employ these strategies, and definitive and rigorous trials are needed to clarify safety and efficacy to guide clinical practice.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjo.2023.36.
Data availability
Data availability is not applicable to this article as no new data were created or analysed in this study.
Acknowledgements
D.M.D. is grateful to Helen Elwell, British Medical Association Library, London, for expert help with literature searches and to Emmanuelle Bostock, University of Tasmania, for substantial input during the initial stages of this project..
Author contributions
D.M.D., A.H.Y., A.R. and V.M. conceived the study. D.M.D., J.K.-G., M.H. and W.M. conducted the study selection and data extraction. D.M.D. led all stages of the manuscript preparation and wrote the original draft. All authors read and provided comments on the final manuscript.
Funding
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Declaration of interest
D.M.D. is a GP Partner in the National Health Service (NHS); he has received fees for presentations, including Royal College of Psychiatrists International Congress Edinburgh 2013 (travel and accommodation only), webinars for general practitioners, PRIMHE (Primary care Mental Health & Education), Network Locums and BMJ Masterclasses; he received an honorarium from Lundbeck for a symposium presentation at the British Association of Psychopharmacology 2019 Summer Meeting on ‘The primary/secondary care interface for treating depression: challenges and future perspectives’, which covered practical ways to improve to help patient care and there was no endorsement of any pharmaceutical treatment or product. J.K.-G. owns shares in AstraZeneca and GSK plc. M.H. is supported by an Australian Rotary Health PhD Scholarship. W.M. is currently funded by an NHMRC (National Health and Medical Research Council) Investigator Grant (#2008971) and a Multiple Sclerosis Research Australia early-career fellowship and has previously received funding from the Cancer Council Queensland and university grants/fellowships from La Trobe University, Deakin University, University of Queensland, and Bond University; he has received industry funding and has attended events funded by Cobram Estate Pty. Ltd., has received travel funding from Nutrition Society of Australia, and consultancy funding from Nutrition Research Australia and ParachuteBH; he has received speakers honoraria from The Cancer Council Queensland and the Princess Alexandra Research Foundation. A.H.Y. is Deputy Editor of BJPsych Open and did not take part in the review or decision-making process of this paper; he is employed by King's College London, is an Honorary Consultant at SLaM (South London and Maudsley) (NHS UK); his independent research is funded by the National Institute for Health and Care Research (NIHR) Maudsley Biomedical Research Centre at SLaM NHS Foundation Trust and King's College London; he has given paid lectures and advisory boards for the following companies with drugs used in affective and related disorders: Astrazenaca, Boehringer Ingelheim, Eli Lilly, LivaNova, Lundbeck, Sunovion, Servier, Livanova, Janssen, Allegan, Bionomics, Sumitomo Dainippon Pharma, COMPASS, Sage, Novartis, Neurocentrx; he is Principal Investigator in the Restore-Life VNS (vagus nerve stimulation) registry study funded by LivaNova, Principal Investigator on ESKETINTRD3004: ‘An Open-label, Long-term, Safety and Efficacy Study of Intranasal Esketamine in Treatment-resistant Depression’, Principal Investigator on ‘The Effects of Psilocybin on Cognitive Function in Healthy Participants’, Principal Investigator on ‘The Safety and Efficacy of Psilocybin in Participants with Treatment-Resistant Depression (P-TRD)’, UK Chief Investigator for Compass; COMP006 & COMP007 studies, UK Chief Investigator for Novartis MDD (Major Depressive Disorder) study MIJ821A12201; grant funding (past and present): NIMH (National Institute of Mental Health) (USA); CIHR (Canadian Institutes of Health Research) (Canada); NARSAD (National Alliance for Research in Schizophrenia and Affective Disorders) (USA); Stanley Medical Research Institute (USA); MRC (Medical Research Council) (UK); Wellcome Trust (UK); Royal College of Physicians (Edin); BMA (British Medical Association) (UK); UBC-VGH (University of British Colombia - Vancouver General Hospital) Foundation (Canada); WEDC (Western Economic Diversification Canada) (Canada); MSFHR (Michael Smith Health Research) (Canada); NIHR (National Institute for Health Research) (UK); Janssen (UK) EU Horizon 2020. M.B. is supported by a NHMRC Senior Principal Research Fellowship and Leadership 3 Investigator grant (1156072 and 2017131); he has received grant/research support from National Health and Medical Research Council, Wellcome Trust, Medical Research Future Fund, Victorian Medical Research Acceleration Fund, Centre for Research Excellence CRE, Victorian Government Department of Jobs, Precincts and Regions and Victorian COVID-19 Research Fund; he received honoraria from Springer, Oxford University Press, Cambridge University Press, Allen and Unwin, Lundbeck, Controversias Barcelona, Servier, Medisquire, HealthEd, ANZJP, EPA, Janssen, Medplan, Milken Institute, RANZCP, Abbott India, ASCP, Headspace and Sandoz (past 3 years). V.M. has received research funding from Johnson & Johnson, a pharmaceutical company interested in the development of anti-inflammatory strategies for depression, but the research described in this paper is unrelated to this funding. V.M. is supported by MQ Brighter Futures grants (MQBF/1 IDEA) and (MQBF/4), by the Medical Research Foundation (Grant: MRF-160-0005-ELP-MONDE) and by the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London. The views expressed are those of the authors and not necessarily those of the funders, the NHS, the NIHR or the Department of Health and Social Care.
eLetters
No eLetters have been published for this article.