Ketogenic diet in epilepsy: an updated review

The history of the KD is very long, dating back to antiquity and biblical times. Fasting was observed to have a beneficial effect on reducing devastating seizures which was reported in the works of Hippocrates (5^th century B.C.) and in the Gospel according to Mark. Interest in the dietary treatment of epilepsy was revived at the beginning of the 20^th century, a time when phenobarbital (PB) and bromides were the only available AEDs. French physicians Guelpa and Marie (1911) were the first to describe the impact of fasting on reducing seizure intensity. In the United States, an osteopathic physician Conclin and Macfadden (1922) wrote the first reports regarding the beneficial effect of fasting on epileptic seizures. In 1921, during an American Medical Association (AMA) convention, Rawle Geyelin presented observations of a group of 30 patients aged 3.5–35 years, that fasted for 20 days and which led to absolute seizure freedom in 87% of patients. Woodyatt (1921), having used the diet in diabetic patients, drew attention to the presence of acetone and beta-hydroxybutyric acid (BHBA) in the blood of both healthy people subject to fasting and those on a low carbohydrate and high fat diet. At the same time Wilder (1921) and then Peterman (1925) from the Mayo Clinic presented a proposal of a diet in which most of the energy came from fat and the metabolic processes arising while on the diet mimicked the metabolic state achieved during fasting. Due to the fact that the diet induced the state of ketosis, which likely plays a role in fighting epileptic seizures, this diet was named the KD. General rules for its application have not changed till this day (Wheless, 2008). By 1930, the KD had been used in 272 children in the USA (A.D.B., 1931) as monotherapy or, in some patients, with added phenobarbital. In 1938, the introduction of phenytoin (PHT), expected to be more effective and easier to use than the diet, perceived as a difficult therapy, hampered its wider spreading. More AEDs were successively introduced (since the 1950s – 26 in total) (Elia et al., 2017) and they have raised hopes for complete seizure control; however, around 30% of patients still do not reap the expected benefits of pharmacotherapy. In the 1980s the diet was taken up once more by professor Freeman and his co-workers from Johns Hopkins Hospital in Baltimore, USA, in a limited number of patients (around 10 a year). As a result, in 1992 they published a report summing up KD’s effectiveness (mostly as add-on therapy) in 58 children with epilepsy (Kinsman et al., 1992). Since the results of the first multi-centre studies – evaluating the efficacy of the KD – were presented at the American Epilepsy Society (1996), interest in this therapeutic method has been growing (number of reports from ca. 100/year in the nineties to over 200 in 2017). The principles for using the KD in children were published in 2009 (Kossoff et al., 2009a). As more insight is gained into its mechanism of action, the KD is increasingly being used in diseases other than epilepsy (Stafstrom, Rho, 2012).

Though there is no doubt as to the efficacy of the KD in refractory epilepsy treatment, its precise mechanism of action remains unknown. In the case of pharmacotherapy, in contrast, the mechanisms of many AEDs are known (e.g. through voltage-dependent sodium channels), which might suggest their potential efficacy in epilepsy treatment; nevertheless, it has not brought about a decrease in the percentage of patients with refractory epilepsy, the underlying mechanism of which cannot be entirely explained either (Rho, Stafstrom, 2012). Given the fact that the KD is effective in many types of epilepsy, it must be acknowledged that its mechanism of action is undoubtedly very complex. During fasting, or when the carbohydrate supply is lowered and the fat supply simultaneously increased, large amounts of free fatty acids (FFA), unable to pass the blood-brain barrier, appear in blood. In liver mitochondria, in the beta-oxidation process, they are converted via acetyl-CoA into ketone bodies: acetone, acetoacetate, and β-hydroxybutyric acid. They are subsequently released into the circulatory system, from where they easily cross the blood-brain barrier. During the application of the KD, the metabolic status accompanying fasting is ‘mimicked’ which means that the cells of many organs, especially of the central nervous system, use fat breakdown products, i.e. ketone bodies, as their main source of energy replacing glucose. The ketone bodies, instead of glucose, become the source of acetyl-CoA for the Krebs cycle. This is a key process for the KD’s action. Moreover, the ketones constitute an essential building block in the process of biosynthesis of cell membranes in the developing brain. As the efficacy of the KD in epileptic seizure reduction is proven, it is clear that the KD impacts on neuronal excitability (Bough, Rho, 2007). Multiple reports indicate that its mechanism of action is particular and different from epilepsy therapies used so far (Rogawski et al., 2016). Among recognised mechanisms of the KD’s action, emphasis is put on the role of caloric restriction, decreased glucose supply and glycolysis reduction, as well as on the direct impact of ketone bodies and free fatty acids (including PUFAs – polyunsaturated fatty acids) on triggering a series of complex biochemical, hormonal and even genetic processes (Masino, Rho, 2012; Danial et al., 2013; Rogawski et al., 2016). Ketone bodies, free fatty acids and limited glucose supply influence both directly and indirectly epileptic seizure control through diverse mechanisms. Available studies indicate that in order to achieve a full anticonvulsant effect, co-occurrence of all three factors above is necessary (Bough, Rho, 2007). For the KD’s action, it is essential to decrease the glucose supply, reduce glycolysis and elevate the blood level of ketone bodies, which represent an alternative ‘fuel’ for cells. In the process of anaplerosis (replenishing missing intermediates participating in the Krebs cycle – TCA) they supply the Krebs cycle, responsible for generating energy (ATP) and indirectly responsible for neurotransmitter production (Rogawski et al., 2016). That reduced glucose and elevated fatty acids cause chronic ketosis – which triggers a series of complex biochemical, hormonal or even genetic processes – has been confirmed. Activation of these processes affects the stabilisation and/or stimulation of cellular metabolism. As a result, it inhibits neuronal dysfunctions accompanying convulsant activity. The scope of the present paper does not allow for a detailed presentation of the numerous interdependent elements (established so far or currently being investigated) that determine the final, indisputably-confirmed effect of the KD on epileptic seizure reduction. The most important ones are, however, worth mentioning. The impact of the KD on mitochondrial function is obvious. It consists in increased biogenesis of these cell organelles, elevated production of energy carrier molecules (ATP) inside them, leading eventually to the build-up of energy reserves in cells. In this process, an essential role is played by PUFAs boosting UCPs (uncoupling proteins) that stimulate mitochondria to produce more energy (Bough et al., 2006). Energy production in cells is conditioned by NAD (nicotinamide adenine dinucleotide). The latest reports point to a probable increase in the NAD+/NADH ratio as a key factor in the KD’s action (Elamin et al., 2017). Moreover, the KD helps to reduce oxidation stress in neurons. Enhanced expression of UCPs induces decreased mitochondrial membrane potential, and consequently reduced ROS (reactive oxygen species) production. A rise in energy production in mitochondria and a drop in the number of free radicals contribute to the prevention of neuronal dysfunction, convulsions or even neurodegeneration (Bough et al., 2006; Greco et al., 2016). Furthermore, reduced glucose and the state of ketosis cause the activation of ATP-sensitive potassium channels (KATP) in neurons which results in reducing their convulsant excitability (Bough, Rho, 2007; Masino, Rho, 2012). The KD also causes changes in metabolism and in the release of neurotransmitters. The synthesis of the inhibiting neurotransmitter GABA is increased, while its breakdown slows down, thus increasing its levels in the cerebrospinal fluid. Besides this, ketosis (BHBA and acetoacetate) directly triggers a decrease in presynaptic release of the excitatory neurotransmitter – glutamate which directly leads to the inhibition of excitatory glutamate-dependent synaptic transmission (Bough, Rho, 2007; Masino, Rho, 2012; Danial et al., 2013; Kossoff, Wang, 2013). Increased concentration of norepinephrine and adenosine, related to restricted glucose supply, is also crucial for the anticonvulsant effect of the KD. Through activation of adenosine A1 receptors, adenosine impacts the ATP-dependent potassium channels mentioned above (Masino, Rho, 2012). Among many aspects unpinning the KD’s action, one that has been recently stressed is the increase in the leptin level. Leptin is an endogenous peptide causing a decrease in synaptic excitability (Rho, Stafstrom, 2012). Attention should be paid to the specific role of PUFAs in the twofold reduction in neuronal excitability, directly and indirectly achieved through the activation of fatty acid receptors, PPARs (peroxisome proliferator-activated receptors) in particular (Bough, Rho, 2007; Masino, Rho, 2012; Kossoff, Wang, 2013).

Many authors emphasise the KD’s influence on inhibiting the mTOR (mammalian target of rapamycin) pathway in the brain (its hyperactivity has a convulsant effect and conduces to epileptogenesis), this is yet another way in which the KD produces an anticonvulsant and antiepileptic effect (McDaniel et al., 2011; Danial et al., 2013; Kossof, Wang, 2013). Ketones (KBs) and their metabolic intermediates (acetyl-CoA, aspartate, PUFAs), among other effects, have a strong influence on the expression of genes linked to cellular metabolism and ROS resistance (Youngson et al., 2017). The period of around 2 weeks that is needed to reveal the anticonvulsant effect of the KD correlates with the time required for alterations to gene expression, mitochondrial proliferation and uncoupled proteins’ upregulation (Maalouf et al., 2009). During treatment with the KD, the glucose level in the blood remains correct, despite its minimum intake when on the diet, owing to the fact that it is produced via gluconeogenesis from amino acids and glycerol released from triglycerides in the beta oxidation process. Thanks to that cells that absolutely require glucose, such as red blood cells, devoid of mitochondria, can satisfy their metabolic needs (Walczyk, Wick, 2017). The KD probably also affects the gut microbiome’s composition by increasing the quantity of bacteria producing short-chain fatty acids (SCFAs), which then pass from the bloodstream into the brain, where they are one of the factors impacting the regulation of seizure activity (Newell et al., 2016). Until recently it was believed that, due to the limited ability of ketones to cross the blood-brain barrier in adults, the KD could not be used in this age group. This assumption has turned out to be incorrect. It has been proven that in disease-related stress, the number of proteins transporting ketones (MCT – monocarboxic acid transporters) to the brain increases (Azevedo de Lima et al., 2014). This is evidenced by the growing number of reports on the KD’s effectiveness in the treatment of adult epilepsy.

In summary, the processes described above, as well as some others not mentioned in the present paper, produce the following results: an increase in neuronal energy reserves (increased mitochondrial biogenesis, increased production of ATP), improved mitochondrial function, a reduction in oxidative stress leading to neuronal damage provoked by ROS and reactive nitrogen species (RNS), the modification of neuronal circuits and cell properties towards reinforcement and normalisation of neuronal function, stabilisation of cell membranes, alteration to the excitability and plasticity of neurons, and the attenuation of inflammatory reactions in nervous tissue. Although the exact mechanism of action of the KD has not been entirely understood, it has been assumed on the basis of previous studies that ketones alter the metabolism of neurons, affect the level of neurotransmitters and regulate the development of neurons (Zhang et al., 2018). KD therapy combines numerous beneficial mechanisms of action, contributing to success in the treatment of epilepsy not only by inhibiting epileptic seizures but also by affecting the course of epilepsy (Boison, 2017). Ultimately, this complex mechanism leads not only to the already known anticonvulsant action of the KD but also produces its neuroprotective, anti-inflammatory and epileptogenesis-inhibiting effects (hence the long-lasting anticonvulsant effect persisting after discontinuation of the treatment) (Maalouf et al., 2009; Danial et al., 2013; Masino, Rho, 2012; French et al., 2017, Chorągiewicz et al., 2010). The clinical effect of the KD cannot be explained with reference to only one mechanism of action because, as in the case of AEDs, it is multifaceted and caused by many factors. Even though the AEDs lead to seizure control in many patients, unfortunately not all patients become seizure-free through pharmacotherapy. The KD makes it possible to control seizures in patients with refractory epilepsy. The therapeutic effect of the KD results from its impact on the metabolism and is completely different from the mechanism of action of the AEDs. Studies on the KD may bring about entirely new therapeutic strategies in refractory epilepsy. Further research on the KD mechanism is necessary (Rogawski et al., 2016).

Many epileptologists consider the KD to be one of the most effective treatments for paediatric epilepsy as add-on therapy, but sometimes as monotherapy (even for first-line therapy in some situations), apart from resective neurosurgery (in cases where neurosurgery is indicated) (Kossoff et al., 2009b).

A beneficial effect of the KD on the EEGs has also been reported, especially on the decrease of interictal epileptiform discharges, just after 1 month of treatment. Patients in whom such a result is visible are more likely to experience seizure reduction after 3 months. Even a short treatment with the KD may cause an increase in beta activity, which suggests that the KD has an effect similar to that of AEDs with a GABAergic mechanism of action (Kessler et al., 2011). Improved background activity is also observed (Li et al., 2013a).

There are a number of disorders in which the KD cannot be used. Inborn metabolic defects are the main absolute contraindications. Patients with previous fat metabolism disorders may present with serious, even life-threatening complications after receiving the KD.

The following errors of fat metabolism are the absolute contraindications to the KD:

In addition, the following disorders constitute absolute contraindications to the KD: pyruvate carboxylase deficiency, porphyria, glycogen storage diseases (except type 2), prolonged QT syndrome or other cardiac diseases, liver, kidney or pancreatic insufficiency, hyperinsulinism. The conditions below are listed among relative contraindications for the KD treatment: inability to maintain adequate nutrition, surgical focus identified by neuroimaging and video EEG monitoring, parent or caregiver/patient noncompliance, and severe gastroesophageal reflux (Kossoff et al., 2009a; Kossoff, Wang, 2013; Sharma, Jain, 2014; van der Louw et al., 2016). The KD should not be used in patients with renal stones or hyperlipidaemia (Vezyroglou, Cross, 2016).

There are no significant contraindications to the simultaneous use of the KD and AEDs. Usually, the diet is added to previously used medicines (except when it is used as the first-line treatment). The prospect of discontinuing medication is generally, in addition to seizure control, the main expectation of patients/caregivers. However, making changes in their administration is not recommended for at least the first 3 months. After this period, if the KD is effective, it is possible to discontinue or reduce AEDs, but there is no numerical data available. Weaning off PB and benzodiazepines is more likely to cause seizure aggravation compared to other medicines (Kossoff et al., 2009a; McArtney et al., 2017).

The influence of the KD on the absorption of AEDs may be significant due to the nausea and vomiting often observed at the beginning, as well as due to the slowed gastric emptying induced by the diet. These conditions may lead to a decreased concentration of AEDs and in consequence to a transient exacerbation of seizures. Sometimes during KD implementation, the concentration of drugs in the blood temporarily increases compared to previous levels (changes in pH, decreased protein intake, decreased urinary excretion) – in the case of PB even up to 100%. Besides that, during treatment with the KD no significant differences in the concentration of AEDs compared to the pre-diet period have been noted (Zupec-Kania et al., 2013; McArtney et al., 2017). Despite many years of combined use of AEDs and the KD, there are only a few reports available on their pharmacodynamic interaction (McArtney et al., 2017). Morrison et al. (2009) analysed a group of 115 children simultaneously treated with the KD and different AEDs: levetiracetam (LEV), LTG, PB, topiramate (TPM), VPA and zonisamide (ZNS). They stated that children receiving PB were significantly less likely to have a <50% seizure reduction than children treated with other AEDs (p = 0.003). This result, as well as the previously described potential increase in PB concentration in the blood, suggest that their simultaneous use should be avoided. The opposite effect occurred in children treated with ZNS – the probability of a <50% seizure reduction compared to those receiving other medicines was significantly higher (p = 0.04) (Morrison et al., 2009). This result may indicate a synergistic effect of the KD and ZNS. The use of the KD in combination with a VNS is another example of a synergistic interaction (Kossoff et al., 2009a).

Taking into account the huge number of patients treated simultaneously with the KD and AEDs, their mutual adverse interactions are rarely observed; however, particular attention should be paid to the concurrent use of the KD and VPA. Ballaban-Gil et al. (1998) noted that most adverse effects, including serious ones such as severe hypoproteinemia, fatty liver disease or Fanconi syndrome, occurred in children who received VPA at the same time as the KD. This was not confirmed in later reports. Other authors did not find a significant difference in adverse effects between children receiving VPA compared to those treated with other medicines (Kang et al., 2004; Lyczkowski et al., 2005). Recently, the problem of the impact of the KD on AEDs used simultaneously has again attracted interest. Heo et al. (2017) analysed 139 patients with refractory epilepsy treated simultaneously with the KD and AEDs as a mono or polytherapy. They compared serum concentrations of AEDs (CBZ, LTG, LEV, TPM, OXC, PB, PHT, and VPA) prior to the introduction of the KD and during treatment. Important differences were found only in the case of VPA – a statistically significant (p > 0.001) decrease in concentration, and PB – an increase in concentration, although not statistically significant, which points to the need to closely monitor their levels after commencing the KD (Heo et al., 2017). Spilioti et al. (2016) observed a rapid increase in ketosis, with clinical symptoms, after discontinuation of VPA in 2 four-year-old girls with refractory epilepsy (out of 73 patients receiving the KD, combined with VPA). Stevens et al. (2016) described the case of an 18-month-old girl receiving TPM and VPA at constant doses, who presented with liver failure symptoms several days after adding the KD. Following her recovery, the KD was used again without VPA, and this time no complications were observed. This particular interaction between the KD and VPA can be put down to common metabolic pathways. VPA competes with fatty acids in the process of beta oxidation in the liver and slows their conversion into ketone bodies. During treatment with the KD, due to the lower binding capacity of proteins, the free fraction of VPA, which may exert a hepatotoxic effect, increases, but excretion of the drug is also higher (Spilioti et al., 2016; Heo et al., 2017). These facts indicate that caution should be exercised when simultaneously using the KD and VPA; additionally, blood levels should be monitored and doses changes as necessary. It should be noted that, drugs that are carbonic anhydrase inhibitors (TPM, ZNS), as well as the KD (especially in the early phase of treatment), predispose patients to metabolic acidosis (a lowered blood concentration of bicarbonates) and also to kidney stones. Therefore, if they are used simultaneously with the KD, it is recommended to monitor the level of bicarbonates and consider their possible supplementation (Takeoka et al., 2002; Zupec-Kania et al., 2013; McArtney et al., 2017).

The KD, like any medical therapy for serious illnesses, may cause adverse effects. Most of them are mild, can be prevented, and if they occur, can be fairly easily treated (Kossoff et al., 2009a). In the early stages, when introducing the KD, the following adverse effects are most common: dehydration, transient hypoglycaemia (>40 mg/dl), hyperketosis, metabolic acidosis, gastrointestinal disorders (vomiting, diarrhoea, abdominal pain, constipation), anorexia, sensation of hunger, exacerbation of symptoms of gastroesophageal reflux, and lethargy. These occur despite the earlier exclusion of contraindications to the use of the KD and are not significantly related to the method of initiation (with or without fasting) although, according to some authors, they are more common in children who endure interruption in feeding, especially >2 years old. They are mainly transient and can be fairly easily remedied. Only rarely are they the reason to discontinue KD treatment (Kim et al., 2004; Kang et al., 2004; Bergquist et al., 2005; Kossoff et al., 2008b; Kossoff et al., 2009a; Luat et al., 2016; Lin et al., 2017; Cai et al., 2017).

A number of ‘distant’ adverse effects that appear only after a few weeks, months or years of treatment with the KD are also known.

Kidney stones may occur in 5–10% of patients (in 25% of patients treated with KD for more than 6 years). They can be prevented via adequate fluid supply and administration of citrate preparations, especially potassium citrate, which increases the excretion of citrates in the urine and reduces hypercalciuria. This action reduces the risk of nephrolithiasis to 0.9%. The risk of developing kidney stones may be slightly higher in patients receiving carbonic anhydrase inhibitors (acetazolamide, TPM, ZNS) (Kossoff et al., 2009a; McNally et al., 2009; Paul et al., 2010; Kossoff, Wang, 2013; Zupec-Kania et al., 2013).

Relatively often, in 14–59% (Kossoff et al., 2009a) of patients treated with the KD, dyslipidemia (hypertriglyceridemia, hypercholesterolemia) is observed. Recently, the problem of lipid profile disorders during the use of the KD – as well as their potential link with cardiovascular diseases – has generated interest amongst many authors. Zamani et al. (2016) assessed the lipid profile in the course of KD treatment in 33 children, finding a significant increase in total cholesterol, LDL and triglycerides (p > 0.001), with very good efficacy (in 63% reduction of seizures was <50%). Cervenka et al. (2016c) evaluated the impact of the MAD on the lipid profile of adult patients with refractory epilepsy. They observed an increase in total and LDL cholesterol (p = 0.01), with the correct triglyceride level after 3 months of treatment, and the absence of cardiovascular or cerebrovascular events. These values normalised naturally during the year and remained at a normal level in patients whose treatment duration even exceeded 3 years. In a prospective study on 38 patients, Azevedo de Lima et al. (2017) assessed the effect of the KD on any increase in triglycerides, total cholesterol, LDL and HDL, but also on apolipoproteins (ApoA-1 and ApoB) and a small LDL subfraction. These molecules have a particularly atherogenic effect on vascular walls. Alterations in the lipid profile in patients with refractory epilepsy result not only from the KD but also from the impact of some AEDs (CBZ, PHT, VPA) favouring ‘dyslipidemia’ by stimulation of cytochrome P450. These observations, without undermining the benefits of the KD in the treatment of drug-resistant epilepsy, indicate the need to monitor not only the classic lipid profile but also its subfractions. Dyslipidemia mostly resolves spontaneously during treatment or after interventions such as reducing the ketogenic ratio, administering carnitine, adding MCT oil to the diet or adding polyunsaturated fatty acids (PUFAs) (Kang et al., 2004; Nizamuddin et al., 2008; Kossoff et al., 2009a; Zupec-Kania et al., 2013; Yoon et al., 2013; Yoon et al., 2014; Cervenka, Kossoff, 2013; Luat et al., 2016; Ułamek-Kozioł et al., 2016; Cai et al., 2017; Azevedo de Lima et al., 2017).

The impact of the KD on the cardiovascular system’s functioning was also examined, and no significant affects were observed. The effect of dyslipdemia on the carotid intima media thickness was not confirmed after 12 months of using KD, although a decrease in its distensibility was observed compared to baseline values. This parameter normalized after 24 months (Cai et al., 2017; Luat et al., 2016). Ozdemir et al. (2016) in a prospective study of 61 children treated with the KD, did not observe any negative impact on cardiac systolic and diastolic functions. The delayed impact of the KD on the cardiovascular system requires further research.

Among the adverse effects of the KD related to the cardiovascular system, a prolonged QT interval is listed (Kang et al., 2004; Hartman, Vining, 2007; Kossoff et al., 2009a, Elia et al., 2017) as is the rare, but serious complication – cardiomyopathy. Their occurrence is associated with selenium deficiency observed during KD treatment (Kang et al., 2004, Cervenka, Kossoff, 2013). Arslan demonstrated this deficiency’s presence after 6 and 12 months of therapy in almost half (49.1%) of 110 patients assessed, however none of them was diagnosed with clinical, electrocardiographic or echocardiographic abnormalities. These results suggest the need to monitor selenium levels, and if necessary, use selenium supplements during treatment with the KD (Arslan et al., 2016). Supplementation during treatment is also required for vitamins, minerals and trace elements, insufficient in the KD, as their deficiency may be the reason for clinical disorders (Kossoff et al., 2009a; Cervenka, Kossoff, 2013; Luat et al., 2016; Cai et al., 2017). The KD, as well as concomitantly used AEDs, especially VPA, increases the risk of carnitine deficiency, clinically manifested by weakness, decreased muscle strength, hypotonia, fatigue, apathy and also anaemia, cardiomyopathy or disorders of liver function. Hence there is a need to monitor its level in the course of treatment. Should the level of free carnitine be reduced, or clinical symptoms of its deficiency occur, supplementation must be introduced. Some authors recommend prophylactic administration of carnitine in patients on the KD (Kang et al., 2004; Kossoff et al., 2009b; Cervenka, Kossoff, 2013; Fliciński et al., 2016).

Long-term use of the KD may also lead to demineralization and increased bone fragility. Groesbeck et al. (2006) reported bone fractures in 21% of 28 children treated with the KD for more than 6 years. Bergquist et al. (2008) confirmed progressive bone mineral content loss during treatment with the KD. Besides inadequate dietary supply of calcium and vitamin D, the cause of these disorders is also associated with acidosis occurring during KD treatment, poorer nutritional status linked with chronic disease and long-term use of medication affecting the calcium-phosphate balance already prior to the diet. These complications can be prevented by calcium and vitamin D supplementation as well as by, during the use of the KD, avoidance/prevention of excessive metabolic acidosis (by the administration of alkalinising agents, especially if the patient receives carbonic anhydrase inhibitors). In the case of longer use of the diet, the need to periodically repeat densitometric tests (DEXA) is stressed (Kang et al., 2004; Groesbeck et al., 2006; Bergquist et al., 2008; Yuen et al., 2017). Slowed growth, observed in a relatively large number of children on the KD, even according to some authors in more than 80%, has not been entirely explained. Younger children are most exposed to this risk. Its occurrence is more likely in the presence of chronic acidosis and insufficient protein intake. Therefore, monitoring the progress of growth in children is essential, especially in those treated with the KD for a long time (Bergquist et al., 2008; Kossoff et al., 2009a; van der Louw et al., 2016; Luat et al., 2016; Cai et al., 2017).

A high-fat diet can also cause uncommon but dangerous complications such as liver dysfunction and pancreatitis. There is a risk of an increased level of transaminases (in general >200 mg/dl), sometimes already at the beginning of the treatment, a fatty liver after about 6 months and the formation of gallstones 12 months into the diet. The association of toxic liver damage with simultaneous use of VPA during the KD and also with carnitine deficiency, is stressed. These complications are usually reversible and do not require discontinuation of the diet. Patients treated with the KD should be monitored regularly in order to detect potential adverse effects on the liver (ultrasound, transaminase). If pancreatitis occurs (a very rare but life-threatening complication), discontinuation of the KD should be considered (Kang et al. 2004; Cervenka, Kossoff, 2013; Arslan et al., 2016; Stevens et al., 2016).

The increased incidence of infection, without clearly confirmed immune disorders, is mentioned among rarely described, although quite common, adverse effects associated with the KD. Generalised severe infections have been sporadically reported (Kang et al., 2004; Cervenka, Kossoff, 2013).

The KD may also be accompanied by hypoproteinemia, occurring in about 10% of treated patients. Its cause, apart from the limited but adequate supply of protein in the diet, and increased gluconeogenesis resulting from the low supply of carbohydrates, is hard to explain (Kang et al., 2004). It can usually be compensated for by increasing the supply of protein in the diet (e.g. by reducing the ketogenic ratio). Reports on single cases of protein-losing entheropathy during KD treatment, sometimes following a difficult course, shed some light on this problem. Abdominal dynamic scintigraphy may be useful in diagnosing the cause of hypoproteinemia occurring in the course of treatment with the KD (Moriyama et al., 2015; Ahn et al., 2017).

Hyperuricemia, hyponatremia, hypomagnesemia, and zinc deficiency, are also mentioned among the biochemical abnormalities that may occur during the use of the KD, but they are easily correctable (Kang et al., 2004; Kossoff et al., 2009a; Cervenka, Kossoff, 2013). The impact of the KD on haematological parameters was assessed in a group of 33 children with refractory epilepsy. A statistically significant increase was found after 6 and 12 months in the haemoglobin, haematocrit, MCV and after 12 months in Vit B₁₂ (Kose et al., 2018).

Although the KD can cause a number of adverse effects, they are not the main reason for interrupting the therapy. Usually, the treatment is discontinued due to a lack of satisfactory efficacy (49.9%), or difficulty in adapting to its restrictions (11%). Adverse effects are mostly mild and easy to manage or can be prevented. The occurrence of more serious, potentially fatal complications, is not more frequent than during the natural course of symptomatic epilepsy in children. Further studies are needed to assess the long-term effect of the KD on health, even many years after the end of the treatment. The potential link between hyperlipidaemia, even transient, and the development of atherosclerosis and cardiovascular diseases later in life should be examined in particular (Cai et al., 2017).

Several recently published studies have attempted to assess the economic aspect of the KD in the treatment of refractory epilepsy in comparison to pharmacological treatment (care as usual – CAU) and the VNS. The cost differences between KD and CAU were not considerable. The VNS, compared to standard treatment and KD, was more expensive over a 12-month period, while over 5 years it was definitely more economically advantageous (de Kinderen et al., 2011; de Kinderen et al., 2015; de Kinderen et al., 2016; Wijnen et al., 2017). In addition, the number of hospital admissions and visits to the hospital emergency department was analysed in a group of 37 children with refractory epilepsy. A 12-month period of treatment with the KD was compared to the 12 months prior to its introduction. The number of visits to the emergency unit decreased by 36%, and there were also 40 % fewer hospitalisations with a 39% reduction in hospital days. As a result, there was a decrease in the cost associated with hospital treatment and emergency department interventions in the group of children treated with the KD (Kayyali et al., 2016). This issue also requires further long-term research on larger groups of patients taking into account the direct and indirect costs of each form of therapy used.