Ketogenic diet in the treatment of epilepsy

Scientific basis
Article section 3 of 12.  Previous  Next

By Marcio A Sotero de Menezes MD and Jennifer L Schoenfeld ARNP

Understanding the ketogenic diet requires familiarity with the mechanisms by which the human metabolism switches its main source of energy from carbohydrate to fatty acids and ketones. The knowledge of these mechanisms is directly useful in the management of children on the ketogenic diet. The following is a summary of reviews of the metabolic endocrine changes associated with fasting and the ketogenic diet (Sankar and Sotero de Menezes 1999; Cullingford 2004):

  • A ketogenic ratio (fat to carbohydrates and protein) of greater than or equal to 3 to1 is necessary to initiate and maintain ketosis.
  • Beta-oxidation is the main pathway of lipolysis for long-chain fatty acids.
  • Ketosis and lipolysis is maintained with a low insulin to glucagon ratio; therefore, the ketogenic ratio needs to be given with every meal.
  • Enteral glucose is a powerful insulin-release stimulus. Relatively high carbohydrate ingestion in a single meal may dramatically reduce or stop ketogenesis.
  • Glucose has a priming effect on insulin release. When high carbohydrate ingestion halts ketogenesis, subsequent amino acid and glucose ingestion will produce a higher insulin release, and ketosis may be slowed for several hours or days. Fasting may be needed for quick ketogenesis return.
  • High serum ketone decreases the glycolytic activity; consistent ingestion of the ketogenic diet may be helpful in maintaining the state of ketosis.
  • High serum ketone concentrations decrease low serum glucose-associated CNS dysfunction. Therefore, ketotic children may not necessitate such aggressive treatment for hypoglycemia (especially asymptomatic) as nonketotic children.
  • High serum fatty acids enhance fatty acid catabolism and ketosis by inducing transcription of the genes for key enzymes involved in lipolysis and ketogenesis.
  • Acidosis may increase the conversion of acetoacetate into beta-hydroxybutyrate. Therefore, beta-hydroxybutyrate may be a more reliable measure for children on the diet because excessive ketosis may produce acidosis.
  • High serum ketone concentrations are associated with a greater chance of seizure control. Ketotic rodents have a higher threshold for electroshock-induced seizure.
  • Children who are less than 12 years old have higher efficiency of extraction of ketones through the blood-brain barrier.
  • Because fasting increases the blood-brain barrier permeability to ketones, ketogenic diet initiation with fasting may be better when a quick effect is desirable.
  • Chronic ketosis increases the blood-brain barrier permeability to ketones. Consistent use of the ketogenic diet allows better results. Inconsistent ketosis tends to produce the worst seizure control, rather than fluctuation in the number of convulsions.

Fatty acid oxidation and ketone body formation. When the carbohydrate intake decreases, the adipocytes start breaking down the triglycerides into glycerol and fatty acids. Subsequently, free fatty acids are released into circulation. The liver and muscles use fatty acids as an energy source. The main process responsible for the lipolysis is the mitochondrial beta-oxidation. The hepatocyte fatty acids can be oxidized into ketone bodies, which can be used as energy substrate by the brain and other tissues, such as muscle.

Beta-oxidation. Beta-oxidation reduces fatty acids sequentially by the repetitive removal of 2 carbon fragments at the carboxy-terminal end. Each cycle of the beta-oxidation produces reduced electron-transfer flavoprotein (electron-transfer factor red), NADH (+) H(+), and acetyl-CoA. The metabolic fate of fatty acids is dependent on the chain size. The larger fatty acids are initially cleaved by their size-specific enzymatic system, and subsequently use the smaller size systems (Schulz 1990; Bale 1992; Haas and Marsden 1996):

  • Very long-chain fatty acid is 22 C to 16 C
  • Long-chain fatty acid is 18 C to 12 C
  • Middle-chain fatty acid is 12 C to 4 C
  • Short-chain fatty acid is 6 C to 4 C

There is some overlap in the size specificity of these systems (Schulz 1990).

There may be also differences in the rate of oxidation between fatty acids depending on their structure. The oxidation of alpha-linolenate is twice that of linoleate or oleate (Cunnane 2004). The decreasing order of beta-oxidation rate is alpha-linolenate > linoleate > oleate > palmitate > stearate (Cunnane 2004). In light of that, it only makes sense that alpha-linolenate is also one of the most ketogenic fatty acids and that enriching a high-fat diet with those compounds increases its ketogenic potential in rats (Gavino and Gavino 1991; Likhodii et al 2000; Cunnane 2004). This is the reason that, in rats, a diet high in flaxseed oil (which contains 60% of alpha-linolenate) is more ketogenic and offers more seizure protection than those using other sources of ketones in rats (Likhodii et al 2000; Cunnane 2004).

Fatty acids greater than 12 carbons in size can only penetrate the outer mitochondrial membrane after being transformed into their acyl-CoA esters. After that, the acyl-CoA esters interact with the carnitine-palmitoyl transferase 1, which attaches carnitine and removes the CoA, producing acylcarnitine. Acylcarnitine is then carried through the inner mitochondrial membrane by the carnitine or acylcarnitine transporter into the mitochondrial matrix. The carnitine-palmitoyl transferase-2 enzyme, which is located in the internal surface of the inner mitochondrial membrane, removes the carnitine and reattaches the CoA residue, forming an acyl-CoA ester again. Carnitine-palmitoyl transferase-1 is the rate-limiting step in the beta-oxidation of fatty acids, and it is the site of action of the lipolysis inhibitor malonyl-CoA (McGarry et al 1977). Malonyl-CoA is formed from citrate derived from the Krebs cycle. Increases in free fatty acids augment the transcription of the CPT1 gene (Cullingford 2004). Fatty acids bind and activate the peroxisome proliferator activated receptor (PPAR) alpha, which is a potent activator of the transcription of several lipolysis and ketogenesis enzymes such as CPT1, HMGCS2, and acyl-CoA synthase (Cullingford 2004). PPAR alpha promotes transcription of these enzyme genes via the peroxisome proliferator response element (PPRE) (Cullingford 2004).

Medium-chain fatty acids cross the inner mitochondrial membrane freely. In fact, octanoate (a middle-chain fatty acid) infusion in fed animals promptly induces ketosis, an effect that is not seen with long-chain fatty acid infusions (McGarry and Foster 1971). The lipolysis requires the action of 1 acyl-CoA dehydrogenase, as well as 3 other enzymes: enoyl-CoA hydratase, 3-hydroxyacyl-CoA dehydrogenase, and 3-ketoacyl-CoA thiolase. The entire lipolysis pathway involves several enzymes and steps, and its full description is beyond the scope of this article, but is summarized below. For more information, the reader is referred to (Schulz 1990; Carpenter et al 1992; Haas and Nyhan 1992; Izai et al 1992; McGarry 1992; Ogilvie et al 1994; DiMauro et al 1996; Haas and Marsden 1996; Shulman et al 1997).

Ketone formation. Ketones, or ketone bodies, include beta-hydroxybutyrate (3-hydroxybutyrate), acetone, and acetoacetate. Even though beta-hydroxybutyrate is not technically a ketone because its ketone group is reduced by a hydroxyl (Fukao et al 2004), it has been traditionally considered to be a ketone and will be referred to as such in this publication. Ketone formation takes place mostly in the liver, and to a lesser extent, in the kidneys, using plasma free fatty acids as substrate (McGarry 1992). At the cellular level, the process takes place inside the mitochondria. Most hepatic ketone body production is released into the circulation to serve as energy substrate to other tissues, such as muscle and brain. Lipolysis produces acetyl-CoA. Acetyl-CoA produced by fatty acid oxidation is primarily channeled to hepatic ketogenesis; in contrast, the acetyl-CoA derived from glucose is directed preferentially to the Krebs cycle (Des Rosiers et al 1991; Fukao 2004). Two molecules of acetyl-CoA combine to form acetoacetyl-CoA. The latter undergoes enzymatic conversion to acetoacetate. The enzyme 3-hydroxy-3-methylglutaryl-CoA synthase, which helps in this conversion, is upregulated in animals taking the ketogenic diet (Cullingford et al 2002). PPAR alpha also promotes transcription of the 3-hydroxy-3-methylglutaryl-CoA synthase (HMGCS2) gene (Cullingford 2004). HMGCS2 transcription is also upregulated during fasting, fat feeding, and low insulin states (Cullingford 2004; Fukao et al 2004). The direction of the interconversion of beta-hydroxybutyrate and acetoacetate depends on the mitochondrial redox state (Marliss et al 1970). Acetoacetate can also be decarboxylated into acetone by a slow, nonenzymatic spontaneous reaction (McGarry 1992).

Peroxisomal beta-oxidation: omega and alpha oxidation. Peroxisomes contain a beta-oxidation pathway served by enzymes that are genetically distinct from the mitochondrial enzymes (Lazarow and de Duve 1976). The pathway controls the cleavage of very long-chain fatty acids (greater than 18 carbons) to hexanoyl-CoA. Under conditions of prolonged fasting, approximately 20% of fatty acids are oxidized by the peroxisomes (Krahling et al 1978).

Omega oxidation occurs in the microsomes and replaces the methyl end of the fatty acids with a carboxyl group, resulting in the formation of dicarboxylic acids (Preiss and Bolch 1964; Haas and Marsden 1996). Dicarboxylic acids are found in the urine when the capacity of the mitochondrial or peroxisomal beta-oxidation is surpassed, such as when fasting or feeding with medium-chain triglycerides, in diabetic ketoacidosis, glutaric aciduria (Haas 1992), or inborn errors of beta-oxidation (Hale and Bennett 1992; Haas and Marsden 1996).

Alpha-oxidation is a process required for cleavage of certain methylated fatty acids. The process involves the alpha-hydroxylation of long-chain fatty acids. A typical example of a fatty acid that initially cannot be cleaved by beta-oxidation is the phytanic acid (McGarry 1992). The same process is used to start the oxidation of shorter fatty acids. The reactions take place in the endoplasmic reticulum, as well as in the mitochondria, and require molecular oxygen, reduced nicotinamide nucleotides, and specific cytochromes.

The control of lipolysis and ketogenesis during fasting and high-fat diets. From the beginning, the ketogenic diet was administered as an attempt to replicate and prolong the metabolic effects of fasting (Wilder 1921; Lennox 1928; Lennox and Lennox 1960). Most of what is known about the control of ketogenesis was learned by the study of chronic fasting in adults. The extent of the similarities between children in states of chronically enhanced lipolysis, such as the ketogenic diet and adult fasting, is not known. More experimental data are necessary to demonstrate the similarities and differences between fasting and the ketogenic diet states. The endocrine and metabolic control of lipolysis is summarized in Table 2.

Table 2. Endocrine and Metabolic Control of Lipolysis and Ketogenesis


Facilitation of lipolysis and ketogenesis

Low insulin to glucagon ratio



High glucagon concentrations (effects may be counteracted by a reactive increase of insulin).

A decrease of acetyl-carboxylase activity causes a decreased conversion of citrate to malonyl-CoA.


A decrease of malonyl-CoA causes an increase in CPT-1 activity, which leads to an increase in mitochondrial entry fatty acyl esters.


An increase in mitochondrial entry fatty acyl esters causes lipolysis, which leads to an increase in ketogenesis (liver).


Low insulin concentrations

An increase of hormone-sensitive lipase (adipocytes) causes an increase in serum free fatty acids, which leads to an increase in ketogenesis (liver).


High growth hormone concentrations (effects may be counteracted by reactive increase in insulin) (Gerich et al 1976; Metcalfe et al 1981).

Increased serum free fatty acids and ketones (Sherwin et al 1983).


Increased insulin resistance and decreased insulin receptor binding (Bratusch-Marrain et al 1982; Rosenfeld et al 1982).


An increase of IGF-1 leads to hypoglycemia (DeGroot 1995).


Overnight increase in ketogenesis (Edge et al 1993).


Mild ketogenic action facilitated by somatostatin (Metcalfe et al 1981).


High ketone concentrations

Decreased pyruvate oxidation; decreased glycolysis.


Facilitation of glycolysis and lipid storage

High insulin concentrations

A decrease in hormone-sensitive lipase (adipocytes) causes a decrease in serum free fatty acids, which leads to decreased ketogenesis (liver).


An increase in glucose uptake (increased glucose transporter-4 in muscle and adipocytes).


An increase of glycolysis (increased pyruvate dehydrogenase activity).


An increase of glycolysis causes increased Krebs cycle activity, which leads to increased citrate.


Increased citrate causes an increase of malonyl-CoA, which leads to a decrease of CPT-1 activity, and then a decrease of mitochondrial fatty acyl esters.

Five hormones are the main controllers of lipolysis and ketogenesis: glucagon, epinephrine, cortisol, growth hormone, and insulin. When there is a decrease in plasma insulin and an increase in norepinephrine or epinephrine and glucagon, an increase in the cyclic adenosine monophosphate is seen in adipocytes, which causes an enhancement of the intracellular hormone-sensitive lipase activity. The latter produces an increase in serum free fatty acids. Ketone bodies and insulin do the cyclic adenosine monophosphate feedback control loop, which causes deactivation in the lipolysis process (McGarry and Foster 1977; Foster and McGarry 1980).

Glucose and lipid homeostasis during fasting state. Insulin is the primary regulator of the glucose uptake (Kahn et al 1987), as well as subsequent storage (Bogardus et al 1984) and metabolism (Mandarino et al 1987). In the CNS, 2 glucose transporters exist that are independent of insulin, glucose transporter-1, which transports across the blood-brain barrier, and glucose transporter-3, which is located in the neuronal membrane (Flier et al 1987; Maher et al 1992).

Human studies using C(13) nuclear magnetic resonance demonstrated that gluconeogenesis and glycogenolysis each contribute about half of the body’s entire glucose production during the first several hours of fasting (Rothman et al 1991; Petersen et al 1996). These experiments suggest a gradual initiation of lipolysis during early fasting.

Table 3. Energy Metabolism Changes Associated with Fasting


First 2 days of fasting

Hepatic glycogenolysis and gluconeogenesis occurs (Rothman 1991).


After 48 hours of fasting

Ketone bodies become the main sources of energy (Owen et al 1967; Ruderman 1976).


Third day of fasting

Peak rates of ketone production are reached (Owen et al 1969; Garber et al 1974).


Fourth to the 21st day of


Serum ketone concentrations continue to increase due to decreased muscle consumption, and there is an increased availability of ketones for brain utilization (Owen et al 1967; Owen and Reichard 1971; Shulman et al 1997).

In patients on the ketogenic diet, the chronic administration of large amounts of fat (greater than 90% of the caloric intake) allows for the continued lipolysis after a period of fasting. A low-carbohydrate and high-fat dietary intake is associated with decreased glucose utilization (DeVivo et al 1978), probably due to the direct inhibitory effect of ketosis in the oxidation of pyruvate. The latter effect is associated with an enhancement of the fat catabolism caused by the increased activity of ketogenic and lipolysis enzymes. These induced an increased circulation of polyunsaturated free fatty acids that allowed for a smooth switch from carbohydrate to ketone support of the brain's energy demands (Cullingford 2004).

The role of glucagon and insulin on lipolysis control. When carbohydrate ingestion is decreased during fasting, the hormone-sensitive lipase inside the adipocytes is activated to release free fatty acids into the circulation. Low serum insulin and high catecholamine concentrations mediate this effect (DeFronzo and Ferrannini 2001). Beta adrenergic receptor stimulation-increasing cyclic AMP produces phosphorylation of the hormone-sensitive lipase (DeFronzo and Ferrannini 2001; Fukao 2004). Increased cyclic AMP also induces the dephosphorylation and displacement of perilipin (the protein that coats the triglyceride droplets) that allows the lipase action (Fukao 2004). Insulin is a potent inhibitor of hormone-sensitive lipase adipocytes by induction of its dephosphorylation via the phosphodiesterase 3B (DeFronzo and Ferrannini 2001; Fukao 2004). At least 1 other enzyme, a nonhormone sensitive lipase, has been demonstrated (Fukao 2004). In the hepatocyte, free fatty acids will undergo lipolysis with subsequent production of ketones. When acetyl-CoA produced by lipolysis is in excess of the Krebs cycle utilization, ketones are formed. Up to 90% of acetyl-CoA produced in the liver may be used in the formation of ketones (Bale 1992). Insulin concentrations are directly proportional to the lipid storage, and inversely related to the lipolytic activity in the adipose tissue (McGarry and Foster 1997).

The insulin to glucagon ratio is one of the main determinants of the metabolic switch from glucose oxidation and fatty acid storage to the lipolysis and ketogenesis mode. High insulin to glucagon ratio will facilitate the former, whereas the opposite will enhance the latter.

Glucagon enhances both glycogenolysis and hepatic ketogenesis when the plasma glucose drops (McGarry and Foster 1977; Foster and McGarry 1980). In humans, the effects of glucagon are dependent on the concomitant, and often reactive, insulin concentrations (Gerich et al 1976). When glucagon concentrations increase (within the physiologic range), but insulin concentrations remain relatively low, an increase in beta-hydroxybutyrate and free fatty acids are seen (Gerich et al 1976). When a low insulin to glucagon ratio is present in the portal circulation, acetyl-carboxylase (citrate to acetyl-CoA to malonyl-CoA) activity is decreased, and malonyl-CoA decarboxylase (cleavage of malonyl-CoA into acetyl-CoA) is enhanced (Fukao et al 2004). This produces a lowering of intracellular malonyl-CoA, which increases the carnitine-palmitoyl transferase-1 activity, promoting fatty acid oxidation and ketogenesis (McGarry et al 1977; Shulman et al 1997). On the other hand, when both glucagon and insulin concentrations are elevated, hyperketosis is not seen (Gerich et al 1976). High insulin concentrations shut down the free fatty acid production from adipocytes, depriving the liver from its substrate for ketogenesis (McGarry and Foster 1977; 1997).

Table 4. Stimuli for the Glucagon Secretion

  • Alanine
  • Arginine (also increases serum insulin secretion)
  • Catecholamines
  • Gastric inhibitory peptide
  • Cholecystokinin

Reproduced from (Karam 1997a).

The ingestion of carbohydrates produces the release of insulin, which stops the release of free fatty acids from the fat deposits, thus, depriving the liver of its substrate for the production of ketone (McGarry and Foster 1977). Stimuli for the insulin release are summarized in Table 5 and are reviewed by Cook and Taborsky (Cook and Taborsky 1997) and Karam (Karam 1997a). The control of the insulin release by glucose is pertinent to patients taking the ketogenic diet. A high basal serum glucose concentration as a priming effect (exposure to high serum glucose followed by return to normal concentrations) can increase the magnitude of subsequent insulin release (for several hours) in response to glucose, amino acids, and other secretagogues, such as enteric hormones in vivo (Cook and Taborsky 1997). This is a possible explanation for the prolonged loss of ketosis seen when patients on the ketogenic diet ingest a carbohydrate-rich food. When this occurs, returning to ketosis may take 24 hours or more, because the insulin response to subsequent carbohydrate or amino acid exposure with the next meals may be greater. This “break” in ketosis often offsets the seizure control for 1 day, or at times, several days. A short period (12 to 24 hours) of fasting may be necessary for a quick return to ketotic state (Livingston 1972).

Table 5. Factors Influencing Insulin Release

Factors causing increased insulin release

  • Carbohydrate-rich meal

- Increased serum glucose
- Enteric factors: GIP, glucagon-like peptide, cholecystokinin
- Parasympathetic innervation (in response to local GI stimulation)

  • Dietary amino acid

- Branched chain (especially leucine)
- Arginine

  • High serum ketones (4 to 6 mmol)
  • Phenobarbital (?) (Lahtela et al 1984; 1986; Karvonen et al 1989; Venkatesan et al 1994).
  • Acetazolamide with low glucose concentrations (Boquist et al 1980)

Factors causing decreased insulin release

  • Low serum glucose
  • Epinephrine greater than norepinephrine
  • Diazoxide
  • K positive channel openers
  • Phenytoin. Clinical significance uncertain (Kizer et al 1970; Herchuelz et al 1981; Lebrun et al 1981; Siegel et al 1982; al-Rubeaan and Ryan 1991)
  • Somatostatin
  • Galanin
  • Acetazolamide with high glucose concentrations (Boquist et al 1980)

Acetazolamide, in association with low glucose concentrations, may increase insulin release, which may be the cause of metabolic acidosis seen during fasting (Boquist et al 1980).

In summary, lipolysis and ketogenesis are facilitated by low concentrations of insulin, which increase the availability of free fatty acids (from adipocytes to the serum), and by high glucagon concentrations, which promote lipolysis with subsequent ketogenesis at the hepatocyte.

The role of growth hormone on lipolysis control is summarized in Table 2. There is some evidence that the normal overnight increase in serum ketones is related to growth hormone concentrations, whereas quicker changes caused by diet or fasting are controlled by insulin concentrations (the lower the plasma insulin, the higher the ketones) (Edge et al 1993).

The control of ketonemia. During fasting, and possibly during high-fat diet ingestion, peroxisome proliferator-activated receptor alpha mediates the adaptive metabolic response (Kersten et al 1998; Cullingford 2004). Thus, both at the level of the mitochondrial entry (CPT1 enzyme) and acetyl-CoA, conversion to acetoacetate-free fatty acids promote their own metabolism. On the other hand, excessively high concentrations of serum ketones may produce metabolic acidosis, so a feedback control also exists. When plasma ketone concentrations reach 4 to 6 mmol/L, free fatty acid mobilization from fat tissues decreases (McGarry and Foster 1997). This modulation is thought to be due to an effect of ketones, causing either increased insulin secretion (Madison et al 1964), or a direct action on the free fatty acids at the level of the adipocytes (Williamson and Hems 1970). When some patients who have been on the ketogenic diet long term “skip” a meal, their ketone concentrations may actually drop, because they are dependent on high concentrations of exogenous lipid intake, due to the free fatty acids output being limited by the high serum ketone concentrations feedback effect.

Physiologic and pharmacological effects of ketone bodies. Ketone bodies are one of the main sources of energy during chronic starvation (greater than 48 hours) (Owen et al 1967; Ruderman 1976) and during the neonatal period (Morris 2005). During fasting, when acetoacetate reaches the brain, it is combined with succinyl-CoA to form acetoacetyl-CoA and succinate; it then enters the Krebs cycle to generate energy through the formation of ATP. The same is true in virtually every other tissue in the body except for the liver, as the enzyme acetoacetate-succinyl-CoA transferase is not present in the hepatocytes (McGarry 1992). Acetoacetyl-CoA may be also converted by beta-thiolase into acetyl-CoA, which enters the Krebs cycle for the production of energy (McGarry 1992). In the target tissues, such as the CNS, beta-hydroxybutyrate is converted into acetoacetate as the concentration of acetoacetate is decreased by its utilization in the energy metabolism. Expanded and more complex roles of the brain ketones have been proposed by Cullingford (Cullingford 2004). It has become apparent that, rather than being a passive “absorber” of ketones, the brain also has a local production (Cullingford 2004). Among the possible roles of this local ketone production, and a major one, appears to be that of astrocytes supporting the energy metabolism of adjacent neurons (Cullingford 2004).

Humans may be more resistant to hypoglycemia when high concentrations of ketone bodies are present in the blood, as shown by the infrequent cases of symptomatic hypoglycemia during the course of the ketogenic diet (Livingston 1972) and fasting (Drenick et al 1972; Livingston 1972). Amiel and colleagues studied volunteers in whom controlled hypoglycemia was induced by combined insulin and glucose infusions (Amiel et al 1991). When the volunteers also received a beta-hydroxybutyrate infusion, the epinephrine, growth hormone, and cortisol output in response to hypoglycemia was significantly lower. This may be partly explained by the efficient ketone oxidation in the brain during hypoglycemia (McGarry and Foster 1997). It is logical that ketones increase the threshold for symptomatic hypoglycemia, as acetoacetate in the brain is transformed into succinate or acetyl-CoA, which enter the citric acid cycle to produce ATP.

Children have a greater capability to extract and oxidize ketones (Persson et al 1972; Dodson et al 1976). This effect is even more pronounced during the state of chronic ketosis, when children seem to have an adaptive mechanism that facilitates even more ketone extraction from the blood into the brain (Kraus et al 1974; DeVivo 1983). An increase in the blood-brain barrier permeability to ketones during starvation has also been noticed (Gjede and Crone 1975). The blood-brain barrier appears to be the rate-limiting step in metabolizing ketones from the blood (Hawkins and Biebuyck 1979). This rate-limiting step is probably due to a saturable monocarboxylic acid carrier mechanism, which facilitates the transport of beta-hydroxybutyrate through the blood-brain barrier (Olendorf 1972). The same author suggests that 2-carbon monocarboxylic, 3-carbon monocarboxylic, or 4-carbon monocarboxylic acids can competitively inhibit the transport of each other by this carrier. The monocarboxylate transporter (MCT1) concentrations in adult rats on a ketogenic diet for 4 weeks was 8-fold greater in the brain endothelial cells and neuropil, compared to rats on a standard diet (Leino et al 2001). Human PET data have shown that the brain utilization of beta-hydroxybutyrate is directly proportional to the serum level (Blomqvist et al 2002).

An enhanced metabolism of ketones (Ruderman et al 1974), or ketogenic diet feeding (DeVivo et al 1978), may in fact decrease the oxidation of pyruvate in the brain, thus, lessening glucose utilization. This effect is probably mediated by inhibition of the enzymes phosphofructokinase, pyruvic dehydrogenase, and alpha-ketoglutaric dehydrogenase (DeVivo et al 1978).

Ketosis tends to be more difficult to induce in patients younger than 1 year old and older than 10 years old (Schwartz et al 1989b). In their book, Freeman and colleagues comment on the fact that infants, at times, may not tolerate or benefit from the diet, due to the inability to maintain ketosis or normal glycemia (Freeman et al 1996). Ketosis has been successfully induced in infants by Vining (personal communication, 1998).

Other effects of ketones have been described. Acetoacetate and beta-hydroxybutyrate may serve as precursors of cerebral lipid synthesis in the neonatal period (McGarry 1992). Sodium butyrate at physiologic concentrations may be associated with enhanced apoptosis (programmed cell death) outside of the CNS (Hague et al 1993).

Serum and urine ketone determination. Ketones can be measured in the blood and in urine. Acetest tablets, Ketostix, and Keto-Diastix are semi-quantitative tests and are based on a nitroprusside reaction (Karam 1997a). The urine acetoacetate and acetone is inferred from the color of a paper permeated with the reagent. Urine beta-hydroxybutyrate is generally not measured in these tests. The ketone measurements are obviously dependent on the urinary concentration (often only specific gravity is actually measured). Urine ketones of 4+ (160 mmol/L) are found on a dipstick when blood beta-hydroxybutyrate concentrations exceed 2 mmol/L (Gilbert et al 2000). If the beta-hydroxybutyrate to acetoacetate ratio is high, the results may be abnormally low. Urine concentration (dehydration) may produce spuriously high urinary ketone concentrations.

Serum measurements are also possible. Commonly, the Acetest tablets are crushed, because plasma and serum do not penetrate well in the intact tablet (Karam 1997a). The appearance of a strong reaction after adding a few drops of undiluted serum indicates a serum ketone concentration of at least 4 mmol/L (Karam 1997a). Further quantification is done by successive dilution, and is generally reported as reactivity per dilution. Similarly, the urine Chemstrip can be used for the same purpose. The urine Chemstrip also uses a nitroprusside reaction, which is able to detect serum acetoacetate concentrations as low as 9 mg/dL and acetone concentrations as low as 70 mg/dL. Because these tests measure only acetoacetate and acetone, it is important to note that beta-hydroxybutyrate (not measured) may, at times, be the predominant ketone in the blood. The beta-hydroxybutyrate to acetoacetate or acetone ratio may be increased, due to alterations of the mitochondrial redox state measured as the NADH to NAD(+) ratio (Marliss et al 1970). This is especially true in cases of lactic acidosis (Marliss et al 1970).

Lactate and pyruvate versus hydroxybutyrate and acetoacetate. Reaction 1 is catalyzed by lactate dehydrogenase, and reaction 2 is catalyzed by beta-hydroxybutyrate dehydrogenase. Both reactions are in dynamic equilibrium, and an excess of NADH(+) H(+) will produce an increase in the beta-hydroxybutyrate production. Such is the case in conditions producing an increase in lactate. Decreased conversion of beta-hydroxybutyrate to acetoacetate and acetone may also occur in cases of ketosis associated with insulin deficiency (Ennis et al 1997).

Quantitative microtests measuring serum beta-hydroxybutyrate are available (Stat-site GDS Diagnostics). These are quick and require only 20 µL of blood. Studies of patients on the ketogenic diet point out that the urine ketones may not be a reliable indicator of the serum values and that seizure control of patients on the ketogenic diet correlates better with serum beta-hydroxybutyrate measurements (above 4 mmol/L) and with urine ketones (Chee et al 1997; Gilbert et al 2000). In a small study the measurement of breath ketones has been found to be a reliable indicator of ketosis in adults consuming ketogenic meals (Musa-Veloso et al 2002). More extensive studies with larger numbers of patients are necessary to confirm these findings.

Fasting will produce serum beta-hydroxybutyrate concentrations of 2 to 5 mm (McGarry and Foster 1997), and the range of values in children on the ketogenic diet is 3 to 8 mm. The diet has a higher chance of being effective when serum beta-hydroxybutyrate concentrations are kept between 5 and 8 mm. Freeman and colleagues found that beta-hydroxybutyrate concentrations in blood correlate with seizure control in children on the ketogenic diet, and the mean level for patients with greater than 90% seizure-reduction after 3 months on the diet was 6 mm (Freeman et al 1998a). The data showing a 1:1 correlation between serum ketones and seizure control have been questioned in animal models and possibly even in humans (Cunnane 2004). Ketone concentrations greater than 16 mm are commonly associated with diabetic ketoacidosis and should probably be avoided by children on the ketogenic diet (McGarry and Foster 1997).

During the course of a day, patients on the ketogenic diet will have a build-up of serum ketone bodies that peaks in the afternoon (McQuarrie and Keith 1927; Schwartz et al 1989b). In contrast, in a regular diet, the highest ketone concentrations are in the morning before breakfast due to the overnight fasting and the higher nocturnal concentrations of growth hormone (McGarry and Foster 1977; Foster and McGarry 1980). Urine ketones follow the same pattern.

The effect of medications on ketosis and carbohydrate metabolism.

Antiepileptic drugs. 2-propylpentanoyl-CoA (valproyl-CoA), 1 of the metabolites of valproic acid, has been implicated in inhibition of the mitochondrial fatty acid oxidation (Li et al 1991; Schulz 1991). This compound probably causes the depletion of free CoA inside the mitochondria. Another postulated mechanism is related to the reversible binding of 3-keto-2-propylpentanoyl-CoA (also a valproic acid metabolite) to 1, or several, of the beta-oxidation enzymes (Schulz 1991). Valproic acid may also interfere with the beta-oxidation of medium-chain fatty acids (Bjorge and Baille 1985). This may be due to a direct action of its 2-n-propyl-4-pentenoic acid (Bjorge and Baille 1985). Clinically, it has not been found that valproic acid significantly interferes with the ketogenesis in children on the ketogenic diet. Valproic acid may increase the risk of side effects in patients on the ketogenic diet (Ballaban-Gil 1998).

Beta-blocking agents. Beta-blocking agents inhibit fatty acid and gluconeogenic substrate release and reduce plasma glucagon concentrations (Karam 1997b). Patients on both beta-blocking agents and a diet low in carbohydrates and protein, or those undergoing fasting, are potentially more susceptible to hypoglycemia with decreased capability of ketogenesis. Beta-blocking agents may also decrease the symptoms of hypoglycemia.

Animal models of the ketogenic diet. Several animal models of the ketogenic diet have been described. The highlights of the studies are summarized in Table 6.

Table 6. Animal Models of the Ketogenic Diet

(Appleton and DeVivo 1974)

  • Increased threshold for electroconvulsive seizures in rats
  • Effects peaked between 8 to 20th day on the ketogenic diet

(Uhlemann and Neims 1972)

  • Increased threshold for maximal and hydration threshold electroshock paradigms in mice
  • No effect on electroshock threshold and pentylenetetrazol-induced seizures

(Hori et al 1997)

  • Decrease after-discharge and seizure threshold (kindling model) for the first 2 weeks of treatment
  • No effect longer than 2 weeks of treatment
  • No difference in the after-discharge and seizure duration at time

(DeVivo et al 1978)

  • Increase in the cerebral energy reserve in chronically ketotic adult rats
  • Ketosis predisposes to inhibition of some glycolytic enzymes
  • Possible increase of hexose transport system
  • No difference in brain pH, water content, or electrolytes in the 2 groups of animals

(Kim et al 1998)

  • Anticonvulsant effects in mice even when they are fed ketogenic diet ad lib
  • Good weight gain on ketogenic diet
  • Beta-hydroxybutyrate concentrations achieved were often less than 2 mm

(Stafstrom et al 1999)

  • In rats made chronically epileptic by administration of kainate, the ketogenic diet was associated with fewer spontaneous seizures and reduced CA1 excitability in vitro

(Bough et al 1999)

  • The efficacy of the ketogenic diet is independent of the degree of ketonemia, but is markedly influenced by ketogenic ratios (more fats versus carbohydrates and proteins) and decreasing weight

In summary, most of the animal models of the ketogenic diet do not appear to reflect the human situation. Most of them represent models of seizure challenge rather than chronic seizure models. One study found that the ketogenic diet induced long-term changes in the hippocampal network excitability in a chronic spontaneous seizures model, for instance, the kainate-induced status epilepticus (Stafstrom et al 1999). It is likely that the degree of ketosis achieved by these models is not sufficient to produce the same robust anticonvulsant effect seen in the ketogenic diet in humans.

Mechanisms of action and in vitro studies of the ketogenic diet. The exact mechanisms of action of the ketogenic diet remain elusive in spite of its being used for more than 70 years. Previously proposed mechanisms, such as negative sodium and potassium balances (Millichap et al 1964), have been disproved by subsequent publications (Huttenlocher 1976; Schwartz et al 1989b). Debakan first noticed elevations in serum lipids in patients on the diet (Debakan 1966). Over the past few years, elevations on polyunsaturated fatty acids have been postulated to have some effect in seizure control; this is based on a few clinical and mostly animal work (Yehuda et al 1994; Voskuyl et al 1998; Cunnane et al 2002; Schlanger et al 2002; Fraser et al 2003). Special arachidonate and docosahexaenoate have been shown to have an anticonvulsant role (Cunnane et al 2002). Longer-chain and unsaturated fatty acids are associated with the anticonvulsant activity (Cullingford 2004). Polyunsaturated fatty acids acting in concert with ketones to exert their anticonvulsant role has been proposed (Cunnane 2004). Cullingford also suggests that another possible mechanism involved is that polyunsaturated fatty acid-induced inhibition of cyclooxygenase 2 caused decreased synthesis eicosanoids, which are suspected to potentiate seizures and kainate-induced hippocampal cell death (Kunz and Oliw 2001; Cullingford 2004). Other anti-inflammatory and neuroprotective actions are indirectly mediated by polyunsaturated fatty acids, including decreased action of transcription factors such as NF-kappaB and AP-1, which cause a decrease in the cyclooxygenase and inducible nitric oxide synthase activity (Delerive 1999).

A study demonstrated that the administration of kainic acid in rodents induces acute seizures and leads to neuronal death and cellular and molecular alterations within the limbic structures. This study found that the ketogenic diet is neuroprotective by inhibiting caspase-3-mediated apoptosis in hippocampal neurons. Thus, early implementation of the ketogenic diet can provide an antiepileptogenic effect and potentially prevent associated learning and memory deficits (Noh et al 2008).

In animals, simple caloric restriction already produces some seizure protection. The effect may be mediated by hypoglycemia or other unknown mechanisms (Greene 2001).

Lowering of pH was proposed by Lennox and Lennox to be one of the important actions of the diet (Lennox and Lennox 1960). In spite of occasional peripheral acidosis, however, the pH measured in the brain does not appear to change during the ketogenic diet, as shown in animals (Al-Mudallal et al 1996) and in humans (Novotny 1997). Local changes in brain pH can be underestimated by whole brain measurements.

Other ideas that have been postulated include GABAergic effects, either indirect or indirect, and changes in the amino acid and neurotransmitter metabolism (Yudkoff and al 2001). About the latter, it is intriguing that, in rodent models, norepinephrine is necessary to produce the diet’s anticonvulsant effect (Szot et al 2001). Even though the latter findings have been confirmed, the effect of the Ketogenic diet seems to depend on other factors (Martillotti and Weinshenker 2006).

The ketogenic diet may exert neuroprotective and anti-epileptogenic properties in addition to the known anticonvulsant effects, which improves its potential to serve as a disease-modifying intervention with implications for diverse neurologic disorders (Masino and Rho 2012; Stafstrom and Rho 2012).

The current and past ideas about the mechanisms of action of the ketogenic diet are summarized in Table 7.

Table 7. Possible Mechanisms of Action of the Ketogenic Diet



Current Ideas (not necessarily proven)

Intracerebral or systemic acidosis (Lennox and Lennox 1960)


• Unlikely to be present in animals (DeVivo et al 1978; Al-Mudallal 1996) or in humans (Novotny 1997)

Negative sodium and potassium balances (Millichap et al 1964)


• Unlikely to be a significant effect (Huttenlocher 1976; Schwartz et al 1989b)

Direct anticonvulsant of hyperlipidemia

(Debakan 1966) or of increased free

polyunsaturated fatty acids

(Cunnane 2004)


• Lipid concentrations are unlikely to be a significant agent of the anticonvulsant action of the diet (Huttenlocher 1976; Schwartz et al 1989b)

Increased “GABA shunt” activity and

increased intracerebral GABA (Nordli and DeVivo 1997); caloric restriction

increases brain glutamic acid

Ddcarboxylase-65 and -67 expression

(Cheng et al 2004)


• Not confirmed in animals (Al-Mudallal et al 1996)


• Present in only 3 out of 6 humans studied by MRS (Novotny et al 1997)

Evidence for the presence of direct

anticonvulsant effect of ketones

• Higher serum beta-hydroxybutyrate concentrations allow better seizure control in humans (Huttenlocher 1976; Freeman et al 1998a)


• Diet more efficient at an age when blood-brain barrier ketone extraction is the most efficient (Persson 1972)


• GABAmimetic effects of beta-hydroxybutyrate, acetoacetate (?) (Nordli and DeVivo 1997)


• Acetoacetate may be anticonvulsant in rabbits (Keith 1931; 1932; 1933); acetone directly anticonvulsant (Likhodii et al 2003)


• Unsustained postsynaptic field potentials in hippocampal slices (Arakawa et al 1991)


• Potentiation of GABAA-mediated inhibitory post-synaptic potentials in hippocampal CA1 neurons studies (Ge and Niesen 1998)


• Local decrease of glucose transporter-1 in the epileptogenic zone


• Decreased glucose transport (Cornford et al 1998b)


• Ketogenic diet re-establishes normal energy metabolism in the epileptogenic zone (?)


Evidence for the lack of direct

anticonvulsant effect of ketones

• Medium-chain triglyceride diet-mediated high serum ketones are not protective in seizure challenge models (Thavendiranathan et al 2000)


• Ketone bodies do not have a direct effect on voltage and ligand-gated channels mediating excitatory or inhibitory neurotransmission in the hippocampus (Thio et al 2000)


Other ketogenic diet effects

Neuroprotection mediated but increasing mitochondrial uncoupling protein (Sullivan et al 2004)


Chronic changes in hippocampal


• Long-term changes in the hippocampal network excitability in the kainate-induced status epilepticus chronic spontaneous seizures model (Stafstrom et al 1999)


Alteration of CNS energy metabolism

• Local decrease of glucose transporter-1, causing decreased glucose transport in the epileptogenic region, leading to impaired energy metabolism, which is re-established to normal concentrations by the ketogenic diet (Cornford 1998a)


• Improved CNS energy metabolism on ketogenic diet shown by MRS (Pan et al 1999).


Hypoglycemia or caloric restriction

• Animals on simple caloric restriction without major ketosis have some seizure protection (Greene et al 2001).


Ketogenic diet anti-seizure effect is

mediated in part by norepinephrine

Discovered by Szot and colleagues and later confirmed by Martillotti and Weinshenker (Szot et al 2001; Martillotti and Weinshenker 2006)

In summary, both clinical and experimental evidence substantiates that ketosis is required for the diet to be effective, even though high concentrations of ketone bodies may be the direct cause of seizure control; evidence accumulated by studies of animal models of the diet speak against that (Thavendiranathan et al 2000; Thio et al 2000). The latter evidence would point to the fact that the ketogenic diet exerts its effects by some other mechanism that is coincident with the timing of the appearance of high concentrations of serum ketones. Further studies are necessary to substantiate these findings.

In This Article

Historical note and nomenclature
Scientific basis
Goals and endpoint
Adverse effects
Clinical vignette
References cited