Low-carbohydrate and ketogenic diets: a scoping review of neurological and inflammatory outcomes in human studies and their relevance to chronic pain

Dietary restriction of carbohydrate has been demonstrated to be beneficial for nervous system dysfunction in animal models and may be beneficial for human chronic pain. The purpose of this review is to assess the impact of a low-carbohydrate/ketogenic diet on the adult nervous system function and inflammatory biomarkers to inform nutritional research for chronic pain. An electronic database search was carried out in May 2021. Publications were screened for prospective research with dietary carbohydrate intake < 130 g per day and duration of ≥ 2 weeks. Studies were categorised into those reporting adult neurological outcomes to be extracted for analysis and those reporting other adult research outcomes. Both groups were screened again for reported inflammatory biomarkers. From 1548 studies, there were 847 studies included. Sixty-four reported neurological outcomes with 83% showing improvement. Five hundred and twenty-three studies had a different research focus (metabolic n = 394, sport/performance n = 51, cancer n = 33, general n = 30, neurological with non-neuro outcomes n = 12, or gastrointestinal n = 4). The second screen identified sixty-three studies reporting on inflammatory biomarkers, with 71% reporting a reduction in inflammation. The overallresults suggestafavourableoutcome onthe nervoussystem andinflammatorybiomarkers fromareductionindietary carbohydrates. Both nervous system sensitisation and inflammation occur in chronic pain, and the results from this review indicate it may be improved by low-carbohydrate nutritional therapy. More clinical trials within this population are required to build on the few human trials that have been done.


Introduction
Adoption of a low-carbohydrate diet (LCD) or ketogenic diet (KD) is gaining popularity as a nutritional therapy for many dysfunctions, including those within the nervous system (1)(2)(3) .Both diets focus on the restriction of dietary carbohydrate.An LCD typically reduces intake to below 130 g per day, with a KD further reducing intake to below 50 g per day to induce nutritional ketosis (4) not necessarily achieved with an LCD.
Ketosis is the elevation of ketones in the blood, with a therapeutic KD generally in the range of 0•5-3•0 mmol/l from a reduction in carbohydrate to <50 g per day (5) .The production of ketone bodies from fat oxidation can influence neurons in several ways: as an alternate fuel source with potential efficiency gains and reduced oxidative stress compared with glucose metabolism, as a regulator of neurotransmitter production and clearance, as a signalling molecule promoting neuroprotection, and as an epigenetic modulator to up-regulate gene expression of antioxidant, mitochondrial and anti-inflammatory functions (4,6,7) .As such, ketones have a dual action, both as a fuel source assisting with bioenergetic stability and a signalling molecule regulating many pathways within the nervous system.
An LCD allows for up to 130 g of carbohydrate per day but does not necessarily produce the same level of fat oxidation and may not result in blood ketone elevation.Despite this, smaller reductions in carbohydrate intake still produce favourable metabolic consequences.These include reducing high blood glucose levels that can sensitise sensory neurons and stimulate pro-inflammatory pathways via receptors and channels on the neuron and microglia (8)(9)(10)(11) .Overall, whether the mechanism is glucose reduction, ketone formation or a synergistic relationship between them, the impact of an LCD or a KD on the nervous system is potentially favourable.
Chronic pain is characterised by increased sensitivity within the nervous system and an inflammatory profile (12,13) and as such may respond to an LCD or a KD.There is limited research on how low-carbohydrate nutrition influences neurobiology in this population.We have recently reported relevant pre-clinical research which identified 170 animal model studies using a ketogenic diet to produce changes within the nervous system (7) .These studies investigated the mechanisms reported to occur with nutritional ketosis and how these alter nervous system function, with fourteen broad mechanistic themes emerging.These included: (a) metabolic, energetic and biochemical changes, (b) restoration of cortical excitability to homoeostatic levels, (c) gene regulation, (d) improved mitochondrial function and number, (e) reduced neuroinflammation, (f) adaptive neuroplasticity, (g) neurotransmitter and synaptic regulation, (h) reduced oxidative stress and (i) activation of signalling pathways.We have also more recently identified a significant and clinically meaningful reduction in reported pain of patients undergoing a whole-food ketogenic diet intervention in a pilot clinical trial (14) .
We lacked the capacity to investigate potential neurological mechanisms in that pilot and as such conducted this scoping review.The primary aim of this scoping review was to identify any clinical trials that report changes in adult nervous system function and report outcomes that may be relevant to drivers of chronic pain.The secondary aim was to more broadly identify clinical trials across any research area that report changes in inflammation in response to an LCD or a KD.As such, this scoping review is a more comprehensive approach than other recent neurological reviews that limited their focus to particular diseases or focused on paediatric populations (1,(15)(16)(17) .It also considers LCD/KD research focused on other body systems or populations (such as metabolic studies or athletes) that report on how the diet influences inflammatory biomarkers.

Protocol
The primary research question for this review was: 'How does an LCD or a KD influence the human adult nervous system?', with the secondary research question being 'Does an LCD or a KD influence inflammatory processes?'.Relevant items from the scoping review protocol and PRISMA-ScR checklist from the Joanna Briggs Institute (18,19) were used as the framework.Studies were reviewed in three phases.Phase one captured all publications reporting a human ketogenic diet that met the criteria for daily carbohydrate intake, research type and diet duration.Phase two sorted the retrieved papers into either research reporting adult neurological outcomes for the primary analysis or allocated the remaining studies into categories based on reported research focus.Research papers on paediatric populations that met the eligibility criteria were also grouped to allow for simple quantification but not further analysed.Phase three reexamined all adult papers from phase two for studies reporting common inflammatory biomarkers.The purpose of this broad approach was to avoid missing studies with a different research focus that also reported a neurological or inflammatory outcome, to establish the breadth of research in the neurological area within the context of all ketogenic diet research, and to quantify the extent of ketogenic research in human adult populations.

Eligibility criteria
For the first screen, studies were included if they met the following criteria: (a) human studies that reported a low-carbohydrate diet (LCD) (≤130 g per day and/or <26% of total energy intake (TEI)) or ketogenic diet (KD) (≤50 g per day and/or <15% of total energy intake) (20) , (b) the LCD/KD was an intervention, (c) the diet lasted ≥14 days to allow for adaptation to the new diet and (d) the study design was prospective (observational prospective cohorts where the KD/LCD was a treatment, or experimental, longitudinal pre-post intervention trials including randomised controlled trials).Full texts were reviewed if the abstract did not contain this information.Studies were excluded if they: (a) were single case reports, (b) were retrospective reviews, (c) were cross-sectional or associational studies, (d) used exogenous ketones rather than diet, (e) were reviews, commentaries, letters or conference papers, or (f) were not in English where the required information was not in the abstract (Fig. 1).
For the second screen, studies were then sorted into three groups: (a) any research on adult populations reporting objective outcomes related to nervous system function, pain, cognitive functioning or psychological outcomes (including mood, anxiety, depression and cognitive processing, but not food cravings or hunger); (b) other adult studies which were categorised by research area: metabolic, cancer, sports/performance, general, gastrointestinal or neurological studies with non-neurological outcomes (for quantification, but did not have data extracted); or (c) paediatric papers (for quantification, but did not have data extracted).Many studies evaluate the effect of a low-carbohydrate or ketogenic diet on child and adolescent development.As this was not the focus of this study, paediatric studies were not included for data extraction to eliminate this confounding variable.We do, however, report the number of studies retrieved.
The third screen searched the two adult groups for any research that reported changes in common biomarkers associated with increased inflammation, including reactive oxygen species (ROS), oxidative stress, C-reactive protein (CRP), tumour necrosis factor alpha (TNF-α), leptin, interleukin (IL)1, 6 or 8, or decreased inflammation (adiponectin, fibroblast growth factor 21 (FGF-21), IL10).A concise summary of the PICOS criteria can be found in Table 1.

Information sources and search strategy
An electronic database search including Medline, EMBASE, Cochrane Library for controlled trials, AMED via OVID, CINAHL via EBSCO, Web of Science and PubMed was carried out on 24 May 2021 and included publication dates from database inception to search date.The search targeted any reference to an LCD or a KD intervention using the terms 'ketogenic', 'Atkins', 'carbohydrate-restricted diet' or 'low-carbohydrate diet'.Additional searches included Google Scholar to check identified articles 'cited by' and 'related articles' links, and reference checks on identified articles with subsequent hand search for these and inclusion if they met the criteria.Retrieved references were downloaded into EndNote reference management software (Endnote X7•7•1, Thomson Reuters 2016) and then imported using Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia).

Study selection and screening
Duplicates were removed, then titles and abstracts were assessed against the eligibility criteria in Covidence by two reviewers independently (R.F. and T.F.).Full texts of identified studies were then screened and categorised by RF.Studies combining paediatric and adult subjects were included if the data identified the adult results of interest separately or if the mean/median age was ≥18 years.Studies included in the second phase were imported into Endnote with their full reference and abstract.Then the inflammatory biomarker search terms were used to extract any study with these terms in any reference field.

Data items
The primary outcome of interest was changes in adult nervous system function (such as excitability, seizure frequency/duration/magnitude, nerve conduction, neuroinflammation, cognitive function or psychological changes).Additional data extracted included: author, study title, year of study, trial type,

Data charting process and synthesis of results
An excel sheet was used to compile extracted data items and generate a table of primary outcomes reported.Primary outcomes were reviewed and summarised by R.F. on the basis of whether the LCD/KD produced positive, neutral or negative outcomes, and whether the comparator diet also had significant positive outcomes.

Results
A systematic database search on  32).A re-screen of all included papers excluding the paediatric populations (n = 588) for the key inflammatory terms retrieved sixty-three studies reporting on inflammatory outcomes (Fig. 1).
Most studies (n = 53, 83%) reported favourable neurological outcomes attributable to carbohydrate restriction.These included: reduction in seizure frequency, improved psychological mood states with reduced negative affect, better cognitive and memory functions, reduced narcoleptic sleepiness, reduced fatigue, depression and rate of multiple sclerosis relapse, reduced structural changes and improved cognitive function in Alzheimer's disease, improved motor function in Parkinson's disease, reduced migraine frequency and severity, reduced pain in osteoarthritis, reduced sympathetic nervous system activation, improved heart rate variability and improved recovery following spinal cord injury.Ten papers reported neutral results (38,46,48,49,52,53,59,63,81,84) and one trial reported a negative effect on mood (45) .Three papers reported also similar or better improvements for the comparator diet (43,44,47) .
Inflammatory markers.The secondary outcome of interest was measures related to inflammation.A second screen of studies meeting the criteria retrieved sixty-three studies reporting on inflammatory biomarkers (Table 3).These were mostly metabolic studies (n = 52) investigating chronic lifestyle diseases such as diabetes, obesity and heart disease (57, . The remining studies included three general population (51,136,137) , two cancer (138,139) , two neurodegenerative disease (61,62) , one auto-immune (140) , one musculoskeletal (78) , one sport/performance (141) and one spinal cord injury (83) .There was a large proportion of RCTs (n = 45), seventeen prospective studies and one case series (138) .Time on the diet ranged from 2 weeks (61,89) to 104 weeks (86,88) (mean 16 ± 20 weeks) with fifty-three studies considered a KD, five mixed models (88,92,96,113,117) and five LCDs (107,109,134,135,140) .The most reported inflammatory biomarker was CRP, followed by TNFα, then various ILs. Studies focued on weight loss commonly reported leptin and adiponectin.
Forty-four studies reported positive outcomes (lowered inflammatory markers) from the diet, with fourteen of these also reporting benefits for the comparator diet (91,95,103,106,108,109,113,116,117,122,127,129,131,134) .
Other findings.There are a substantial number of interventional human KD and LCD trials in other research areas reported in the literature that meet the carbohydrate intake criteria and exceed 2 weeks in duration.In addition to the 61 neurological trials, other publications retrieved in the search included: 393 metabolic, 51 sport/performance, 33 cancer, 30 general, 12 neurological (reporting non-neurological outcomes) and 4 gastrointestinal (Supplementary Appendices 2-7).

Discussion
The primary aim of this scoping review was to review adult human studies that report outcomes related to the nervous system in response to an LCD or a KD.The secondary aim was to review the research that reports on changes in inflammatory biomarkers and evaluate the relevance of these findings to chronic pain.The overall results (83% of studies) suggest a favourable outcome on the nervous system from a reduction in dietary carbohydrates.Most studies utilised a KD below 50 g carbohydrate per day (fifty-eight out of sixty-four focused on neurological outcomes (91%), and fifty-five out of sixty-three focused on inflammatory biomarkers (87%)).The more moderate approach to carbohydrate restriction of the LCD also reported positive outcomes, with three out of four LCDs and both 'unclear' diets reporting improved neurological outcomes.Similarly, three out of four LCDs and four out of five mixed protocols also reported favourable reductions in inflammatory biomarkers.
The KD is reported to reduce nervous system excitability through ketone signalling.The actions include: activation of various ion channels (such as ATP-sensitive K þ channels) where ketones reduce excitability, modulation of neurotransmitter levels via the reduction of excitatory glutamate and/or increase in inhibitory GABA (or changes in clearance rates), and improved mitochondrial respiration and number with reduced reactive oxygen species formation (142,143) .Most studies focused on cortical excitability (twenty-one of twenty-two publications) and reported improvements in seizure measures.Our recent scoping review of animal research using KDs (7) identified fifty-three studies reporting on cortical/neuronal excitability, with forty-two reporting improvements that shifted nervous system excitability back towards homoeostatic levels.Increased neuronal excitability (sensitisation) is common in chronic pain presentations (144- 146) and as such may also respond positively to the mechanisms outlined in the review.Additionally, the use of anti-seizure/antiepileptic medication (such as pregabalin) is commonly prescribed for neuropathic pain, providing further evidence for common pathways being involved in both chronic pain and seizure disorders (147) .Given that poor nutrition potentially triggers mechanisms responsible for driving increased nervous system 306 R. Field et al.  (171) ) 5 g added per week De Luis (97) 2016 The ratio of pro-resolution/proinflammatory lipid markers was increased with DHA supplementation increasing AI effect.(positive) Ebbeling (98) 2018  (172) ) Forsythe (100)  Inflammatory risk (arachidonic acid: eicosapentaenoic acid ratios in plasma phospholipids) were more adversely affected by the KD.
KD has also been reported to reduce inflammatory loading, which is supported by the results from this review.Increased levels of inflammatory cytokines (such as CRP, TNF-α and ILs 1, 6 and 8) have been demonstrated in chronic pain presentations, with a potential link between an increase in low-grade inflammation and the development of persisting pain (12,149,150) .Within the nervous system, pro-inflammatory mediators can activate the microglia, increasing their number and size, altering morphology and increasing receptor expression.This results in further upregulation of neuroinflammation leading to maladaptive synaptic plasticity and central sensitisation (151)(152)(153)(154) .In contrast, the presence of ketones and/or the suppression of glycolysis positively influence the microglia, shifting them towards anti-inflammatory phenotypes (155) .In addition, ketone bodies acting as signalling molecules can mitigate NOD-, LRR-and pyrin domain-containing protein 3 (NLRP3) inflammasome activity in the microglia and central nervous system, and act as a ligand for G-protein receptors becoming class 1 and 2 histone deacetylase (HDAC) inhibitors within the brain up-regulating anti-inflammatory and antioxidant pathways (6,155) .Inducing ketosis via a KD to mitigate inflammatory processes can be supported mechanistically (7,156) as well as clinically from the human studies retrieved (Table 3).
All seven migraine studies utilised a KD and reported positive outcomes.There are several mechanisms relevant to migraine development that are influenced by a KD.In addition to cortical excitability and inflammation already discussed, an energy deficit from impaired glucose metabolism/transport, mitochondrial dysfunction and oxidative stress are also implicated in migraine as well as other neurodegenerative disorders (157,158) .In the context of glucose restriction, ketones are supplied to the neuron for fuel from the periphery through the blood-brain barrier, but also by fatty acid oxidation occurring in within the astrocyte (158) .Ketones can meet the energy shortfall and provide bioenergetic stability.They also reduce oxidative stress by up-regulating antioxidant defence and improved mitochondrial respiration (4,159) .Whilst the presence of ketones on a KD provides a plausible mechanistic explanation for pain reduction, the absence of ketones on an LCD does not necessarily remove the potential benefit.
An LCD which does not achieve significant levels of blood ketones may also improve chronic pain through mechanisms related to lower blood glucose levels rather than ketone-mediated actions.Chronic hyperglycaemia results in the production of advanced glycation end products which activate receptors on the neuron, potentially damaging it and promoting inflammation (such as in diabetic neuropathy) (10) .It also results in hyper-insulinaemia where some cells fail to respond correctly to insulin signalling (insulin resistance) linked to the development of chronic pain such as fibromyalgia (160) .A reduction in glucose also reduces the NADH:NAD þ ratio, which signals a reduction in the transcription of pro-inflammatory genes in the microglia (11) .Interestingly, cross-sectional studies investigating the effect of glucose-lowering medication such as metformin report reduced musculoskeletal pain (161) .Overall, the beneficial outcomes on pain from an LCD or a KD may be a synergistic action of lowered blood glucose and increased blood ketones.
Defining the most beneficial diet for chronic pain has been the target of recent systematic reviews (162)(163)(164) ; however no clear diet stands out (162) .Only two studies using a low-carbohydrate approach specifically for pain were found during this review (77,78) .Both demonstrated improvements in pain outcomes, one using a KD (78) and the other an LCD (77) ; however, neither presented ketone levels or measured dietary carbohydrate data.A third study (14) published after the scoping review date also reported significant improvements in pain outcomes using a KD protocol with a significant but small rise in ketone levels and measured daily carbohydrate intake reduced to 70 g per day in the intervention group.The human trials presented support the use of a KD for a reduction in nervous system sensitivity; however, an LCD also reports benefits.It is unclear from the current research what level of carbohydrate restriction is required, whether the effect increases proportionately with carbohydrate reduction, or what diet duration length is required for favourable impacts on nervous system sensitisation.Further research using larger participant numbers is required to help answer these questions which are relevant for clinical application.
A limitation of this review is the confounding variable of weight loss.A large portion of included studies reported significant weight loss, which is also reduces inflammatory cytokines and may be also responsible for positive outcomes.A further limitation was the exclusion of extensive paediatric studies that may have provided further mechanistic evidence (such as Napolitano and colleagues (165) who were the first to show an increase in the antioxidant glutathione in human rather than animal models) (Supplementary Appendix 1).Despite this, an LCD or a KD is supported by mechanistic animal research as well as the human data presented here to provide plausible rationale as to how physiology might be influenced in a way to reduce pain perception.

Conclusion
This scoping review of LCD and KD research identified a large body of prospective adult human dietary intervention trials that reduced carbohydrate intake to below 130 g per day and exceeded 2 weeks in length.From these, there were sixty-four studies reporting neurological outcomes of which 83% showed improvement, and sixty-three studies reporting inflammatory biomarkers of which 71% improved.Both nervous system sensitisation and inflammation occur in chronic pain and as such may be improved by low-carbohydrate nutritional therapy.More clinical trials within this population are required to build on the few human trials that have been done.
Low-carbohydrate and ketogenic diets 297 https://doi.org/10.1017/S0954422422000087Published online by Cambridge University Press diet description, enrolment/completion numbers for the LCD/KD, comparator diet if applicable, age of participants and weight loss.These data items were repeated for inflammatory biomarkers.
24May 2021 retrieved 7876 publications after the removal of duplicates.Title and abstract screen extracted 1548 papers that were potentially eligible, with full text review excluding 701.The included studies were sorted into publications reporting a neurological outcome (n = 64), other adult research with a different focus (n = 524: metabolic n = 394, sport/performance n = 51, cancer n = 33, general n = 30, neurological with non-neuro outcomes n = 12, or gastrointestinal n = 4), or paediatric papers (paediatric neurological n = 227, other n =

Table 2 .
Reported neurological outcomes from human studies utilising low-carbohydrate (LCD) or ketogenic (KD) diets n = 64

Table 2 .
(Continued ) Carbohydrate presented as either (a) ratio of fat: protein þ carbohydrate, (b) grams per day or (c) a daily percentage.