If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Drug-resistant epilepsy (DRE) occurs in a third of patients.
Autonomic- and chronotherapy-associated parameters that contribute to the degree of response to therapy.
A personalized-based algorithm built on epilepsy-related signatures, autonomic signals, and chronotherapy is presented for overcoming DRE.
Despite progress in the development of anti-seizure drugs, drug-resistant epilepsy (DRE) occurs in a third of patients. DRE is associated with poor quality of life and increased risk of sudden, unexplained death. The autonomic nervous system and chronobiology play a role in DRE. In the present paper, we provide a narrative review the mechanisms that underlie DRE and characterize some of the autonomic- and chronotherapy-associated parameters that contribute to the degree of response to therapy. Variability describes the functions of many biological systems, which are dynamic and continuously change over time. These systems are required for responses to continuing internal and external triggers, in order to maintain homeostasis and normal function. Both intra- and inter-subject variability in biological systems have been described. We present a platform, which comprises a personalized-based machine learning closed loop algorithm built on epilepsy-related signatures, autonomic signals, and chronotherapy, as a means for overcoming DRE, improving the response, and reducing the toxicity of current therapies.
]. An estimated 50 million people have been diagnosed with epilepsy worldwide, and the incidence is 16–51 new cases per 100,000/year. Despite the increased use of anti-seizure drugs (ASD), drug-resistant epilepsy (DRE) remains uncontrolled in a third of patients. DRE often persists even after treatment with two or more drugs [
In the present paper, we present a narrative review of some of the possible mechanisms underlying DRE, describe the role of the autonomic nervous system (ANS) in this disease, and summarize the data on the potential role of chronobiology in epilepsy.
A platform, which comprises a personalized-based machine-learning algorithm comprising epilepsy-related signatures, autonomic signals, and chronotherapy, is described as a potential approach for overcoming DRE and improving the response to therapies.
1.1 The problem of drug resistant epilepsy: etiology and current methods for overcoming it
The International League Against Epilepsy (ILAE) characterized DRE based on two levels [
]. Level 1 categorizes the outcome of each therapeutic intervention as either "seizure free" or as management failure. This scheme considers that the patient was treated “adequately,” meaning that the ASD was appropriately chosen, well tolerated, and appropriately administered. Level 2 defines DRE as a failure of two treatments, either as monotherapies or as combination regimens [
]. The target hypothesis suggests a change in cellular targets of ASD, such as compositional alterations in voltage-gated ion channels and neurotransmitter receptors, which reduce the sensitivity to treatment [
]. The neural network hypothesis for a mechanism of epileptogenesis suggests that degeneration and remodeling of neural networks caused by seizures inhibits the drugs from accessing their neuronal targets. The molecular basis of this concept was described in patients with TLE, who form new excitatory circuits because of progressive sprouting. Neurogenesis and astrogliosis contribute to the development of abnormal neural pathways leading to drug resistance [
The multidrug transporter hypothesis is based on the premise that there is an overexpression of efflux transporters in the liver, gut, and kidney that reduces plasma levels of ASD, decreasing the amount that reaches the epileptic focus [
]. Epilepsy patients with partial responses showed lower plasma levels of phenytoin compared with patients with complete responses, regardless of P-gp levels, and this effect was independent of the dose of phenytoin [
]. On the other hand, a randomized controlled study of 1847 epilepsy patients, which compared immediate vs. deferred treatment with ASD, showed that immediate treatment reduces seizure recurrence in years 1–2; however, long term remission was similar in both groups [
]. Only a portion of DRE cases have a genetic etiology.
Treatment options for DRE are ASD, surgery, neuromodulation, a ketogenic diet, and cannabis. ASD is selected based on the basic principles of epilepsy, and on sex, fertility, age, body weight, drug interactions, and comorbidities. Pharmacological synergistic effects generated by the combination of valproic acid (VPA) and lamotrigine, or the addition of subsequent drugs has a small likelihood of inducing remission [
]. Controlled clinical trials in adults with medically intractable focal seizures treated with responsive neurostimulation (RNS) showed that closed-loop responsive neurostimulation to the seizure focus reduces the frequency of disabling seizures. Median seizure reductions of 75 % after 9 years of treatment was demonstrated [
]. A ketogenic diet (a high-fat, low-carbohydrate diet) results in urinary ketosis that simulates starvation, while maintaining caloric demand, has been associated with a 50 % reduction in seizures in children with DRE [
]. A recent meta-analysis of observational clinical studies on the treatment of refractory epilepsy with cannabidiol (CBD)-based products reviewed 11 studies describing the results in 670 patients. Two thirds of patients reported improvement in the frequency of seizures, and a higher improvement rate was described in patients treated with CBD-rich extracts than in those treated with purified CBD. Thirty-nine percent of responders reported a "50 % reduction or more in the frequency of seizures" although there were no differences between the two compounds [
Taken together, the high variability between studies suggests multiple mechanisms underlying DRE, making it difficult to control. Increasing drug dosages, changing drugs, brain stimulation, or brain surgery are some of the measures taken in these patients, but with only moderate success [
]. The benefits of combination therapy with two or more ASD have been difficult to establish. Rational polytherapy, which combines ASD with differing mechanisms of action in the hope of improved efficacy, has not been consistently proven effective [
]. Therefore, there is a need to find ways to overcome adaptation to ASD.
1.2 The role of the central autonomic network in epilepsy and in drug resistance epilepsy
The central autonomic network (CAN) comprises the amygdala, anterior insula, anterior cingulate cortex, and orbitofrontal cortex. These regions associate with subcortical regions of the CAN, including the hypothalamus, periaqueductal gray, parabrachial areas in the pons, solitary tract nucleus, and ventrolateral medulla [
]. Inputs into the CAN include viscerosensory feedback transmitted via the solitary tract nucleus, as well as humoral inputs transmitted via circumventricular parts. The CAN incorporates visceral, humoral, and environmental data and coordinates endocrine, autonomic, and behavioral responses to various triggers. CAN activity is state dependent, and it is also affected by the sleep-wake and circadian cycles.
Epileptic networks are intimately connected with the CAN. Autonomic symptoms and signs are a common occurrence in seizures, although they are often overshadowed by dominant motor phenomena [
Several studies have examined the effects of epilepsy on the ANS, through quantifying autonomic functions in patients with epilepsy, both in the ictal/peri-ictal and interictal states, with the autonomic cardiovascular regulation being the most studied. Heart rate variability (HRV) is an indirect measure of the ANS influence on the heart and generally considered to reflect sympatho-vagal balance. In general, an increased HRV reflects a shift toward parasympathetic dominance, whereas a decreased HRV indicates a relative increase in sympathetic activity.
Most HRV studies have used interictal electrocardiography data. Two meta-analyses showed that despite high heterogeneity between studies, possibly due to diverse sample characteristics, adult patients with generalized epilepsy compared to healthy, age-matched controls, manifest disturbed interictal autonomic function. Twenty-four-hour HRV and awake measures, comprising variance, high frequency (HF), low frequency (LF), and LF/HF ratio, are reduced, signifying a shift toward sympathetic dominance [
]. A few studies have examined HRV in the peri-ictal state, showing inconsistent results, but there also seems to be a general trend toward sympathetic predominance in all kinds of seizures, most prominently TLE and generalized seizures [
The ANS in epilepsy is controlled by brain-derived neurotrophic factor (BDNF) and insulin-like growth factor 1 (IGF-1). Chronic or severe epilepsy and, mainly, TLE may disturb BDNF and IGF-1 signaling in the central ANS, leading to autonomic dysfunction and reduced cerebral autoregulation in subjects with focal epilepsy. During the interictal period, lowered levels of BDNF and IGF-1 are associated with decreased autonomic functions and reduced cerebral autoregulation [
Sudden, unexpected death in epilepsy (SUDEP) is an important reason for mortality associated with epilepsy. The incidence of SUDEP is increased in patients with DRE. Autonomic dysfunction is associated with sudden death irrespective of structural heart disease. Respiratory and cardiovascular dysfunction were proposed as a mechanism for SUDEP [
]. Subjects with epilepsy manifest irregularities in both sinoatrial node pacemaker and ventricular repolarizing currents, increasing their risk for serious cardiac arrhythmias. Cardiac arrhythmias in patients with DRE can explain SUDEP in these patients [
]. Most cardiac autonomic and ventricular function irregularities are noted during ictal and postictal periods; interictal epileptogenic activity also promotes autonomic imbalance. Patients with untreated recently identified epilepsy showed alterations in HRV that reflect the sympatho-vagal imbalance during the interictal period, including reduced time domain measures, mean HF values, and an increased mean LF and mean LF/HF parameters, along with subclinical deterioration of left ventricular functions, which are associated with increased cardiac mortality [
]. In a single trial, patients with epilepsy showed increased left ventricle stiffness and left ventricle filling pressure, and increased left atrial volume and parameters of ANS dysfunction, including reduced chronotropic index and predicted peak heart rate levels during effort. Autonomic dysfunction can account for 52 % of the stiffness observed [
The vagus nerve (VN), a main component of the ANS, also regulates the neuro-endocrine-immune axis via stimulation of the hypothalamic-pituitary adrenal axis and central ANS, and the efferent cholinergic anti-inflammatory pathway (CAP). The CAP exerts an anti-TNF effect via the discharge of acetylcholine at the distal VN and connection of the VN with the spleen through the splenic sympathetic nerve [
]. VNS does not alter the parasympathetic cardiovagal tone. No differences have been described in markers of parasympathetic cardiovagal tone, or baroreflex sensitivity, between baseline, a 6-month visit, and a final visit. Systolic and diastolic blood pressure (BP) upon 5-min of head-up tilt was increased following VNS implantation [
]. VNS of vagal afferents at increased frequency (20−30 Hz) is beneficial in DRE. Low-frequency (5 Hz) VNS of vagal efferent activates the CAP, leading to an anti-inflammatory effect in inflammatory diseases [
]. In a study of 21 epilepsy patients, VNS affected both sympathetic and parasympathetic cardiovascular modulation without a negative effect on autonomic cardiovascular regulation. VNS did not change respiratory rate intervals (RRI) and BP values. The LF and HF power of RRI and the LF power of BP were augmented, and the HF phase between RRI and respiration decreased significantly [
]. Patients with DRE who were undergoing VNS implantation showed a reduced frequency domain (VLF, LF, HF, TP), time domain (SDNN, RMSSD, pNN50), and nonlinear (SD1, SD2) HRV parameters. Non-responders manifested reduced cardiac autonomic function. Pre-surgical HRV parameters, representing parasympathetic cardiac regulation or vagal tone, correlated with responsiveness to VNS [
]. In subjects with generalized seizures, augmented parasympathetic activity to values above those prior to the seizure were shown. These were followed by a reduction to normal values subsequent to the seizure [
]. In a study of 167 patients with epilepsy associated with unilateral mesial temporal sclerosis, 735 seizures were evaluated. Higher occurrences of seizures were noted during the periods 08:01-12:00 and 16:01-20:00, with fewer seizures between 0:01 and 4:00 [
]. In a large analysis of patients (SeizureTracker cohort), 891 out of 1118 patients (80 %) showed circadian modulation in their seizure rates. This was also observed in 11 out of 12 patients of the NeuroVista cohort. In the SeizureTracker cohort, 7–21 % of 1118 subjects manifested robust circaseptan (weekly) rhythms, with a 7-day period [
]. The association of seizure timing with fluctuating rates of interictal epileptiform discharges, the interictal epileptiform activity (IEA) that is a marker of brain irritability, was observed among seizures. In a further study, IEA oscillated with circadian and subject-specific multi-day periodicities. The multidien periodicities were most frequently 20–30 days, and seizures occurred especially during the rising phase of multidien IEA rhythms [
]. Seizures can follow a 24 -h non-random or non-uniform shape. In a study of 544 seizures in 123 consecutive subjects, the seizure-specific times were spread along 3- or 4 -h time blocks during a 24 -h period. Non-uniform distribution of seizures was noted in temporal lobe epilepsy, showing two peaks found in both 3- and 4 -h periods. However, as peak times vary between studies, rhythmic exogenous triggers or environmental/social “zeitgebers” may modulate the 24 -h rhythmicity of seizures [
]. Generalized seizures occur most frequently out of sleep and in older patients. In a study of 71 patients with a total of 223 seizures, sleep/wake seizure spreading predicted tonic-clonic development better than the time of day, with more seizures occurring during sleep. Tonic-clonic seizures are more frequent between 00:00-03:00 and 06:00-09:00 [
Patients with DRE manifest reduced time domain, frequency domain, and non-linear domains of HRV parameters. Inhibition of cardiac autonomic modulations is more pronounced during the night. Most differences in HRV values are noted in the early morning (usually 05:00 or 06:00), corresponding to cardiac autonomic dysfunction and subsequent SUDEP, which occurs during night or early morning [
]. Mesial TLE (mTLE) is characterized by a recurring pattern of spontaneous seizures, suggesting a reliance on the endogenous clock. Seizures can interfere with the biological rhythm output including circadian oscillation of body temperature, EEG pattern, locomotor activity, and the transcriptome. Dysregulation of the circadian clock in the hippocampus combined with multiple uncoupled oscillators is associated with the development of seizures [
]. mTLE seizures peak in the late afternoon and early evening. Both excitatory and inhibitory alterations in the circadian functions of the dentate gyrus (DG), which controls the generation of limbic seizures, add to this circadian rhythm. In an animal model of mTLE with chronic epilepsy, DG excitability is higher in the afternoon and early evening, underlying the time of day-dependency of seizures [
The intergeniculate leaflet (IGL) of the thalamus regulates the circadian rhythm, and its network is vastly GABAergic, consisting of neuropeptide Y-synthesizing and enkephalinergic neurons. Putative enkephalinergic IGL neurons generate action potentials only during light-on conditions. Absence epilepsy (AE) is characterized by spike-wave discharges in the EEG, induced by hypersynchronous thalamo-cortical oscillations. In an animal model, reduced GABAergic synaptic transmission in the IGL, with a higher firing rate of infra-slow oscillations (ISO) and an irregular response to alterations in continuous lighting, have been reported [
]. Disturbed light detection mechanisms in AE support sleep-promoting system insufficiencies and other arousal disturbances seen in these patients. Malfunctioning of the light recognition system from the retina to subcortical visual assemblies, which is dependent on oscillatory functions, has been suggested in patients with AE. In a rat model of AE, altered rhythmic function of the lateral geniculate neurons, with an increase in both the infra-slow and fast oscillatory frequencies, has been shown. An altered reaction to sustained changes in ambient light supported alterations in the IGL neuronal firing [
]. In a rat model of TLE, after injury by SE, a persistent phase shift of approximately 12 h emerges and development of chronic spontaneous seizures proceeds. Locally, an impaired circadian input to the hippocampus induced spontaneous hippocampal EEG spike (SPK) phase shift. Alterations in the powers of circadian input produced a phase shift in hippocampal neural function. A stable circadian input is required for maintaining natural circadian phases in the hippocampus. Injury to circadian centers, such as the medial septum, which interrupt this equilibrium and alter the circadian regulation, can induce daily rhythms of seizures [
]. Sleep patterns in adults with epilepsy (AWE) are irregular if seizures are uncontrolled or happen during sleep. Obstructive sleep apnea is more common in AWE who are older, male, obese, or whose seizures are uncontrolled [
]. Sleep deprivation (SD) increases the incidence of interictal epileptiform discharges (IED) compared to TLE. SD increases the instability of morning recovery sleep compared with night polysomnography (n-PSG). An increased instability of morning recovery sleep augments IED yield in SD-EEG in TLE patients [
The sleep architecture of a Drosophila voltage-gated sodium channel (VGSC) gene mutant that harbors a human generalized epilepsy with febrile seizures plus (GEFS+) mutation suggests a relationship between sleep and epilepsy [
]. GEFS+ mutant sleep phenotypes are unaffected by pharmacologic reduction of GABA transmission by CBZ and are alleviated by decreasing GABA receptor expression in wake-promoting pigment dispersing factor (PDF) neurons. The data suggests that an elevated GABAergic transmission to PDF neurons accounts for increased nighttime sleep in GEFS+ mutants. [
]. In a rat model, kindled epilepsy was induced at three different ZTs, and sleep-wake functions were analyzed prior to and following a seizure. Expression of PER1 protein in the hypothalamic SNC and the circadian rhythm of sleep instability were advanced by a few hours upon stimulation at ZT6. Changes in sleep circadian rhythm and PER1 oscillation, prompted by ZT6-kindling, were inhibited by a hypocretin receptor antagonist in the SCN. ZT6-kindling stimuli changed the circadian oscillator. ZT0-kindling decreased rapid eye movement (REM) and non-REM (NREM) sleep, mediated by the corticotrophin-releasing hormone. ZT13-kindling increased interleukin-1 and subsequently augmented NREM sleep without changing the sleep-wake fluctuation [
Differences in chronotypes among patients with different epilepsy syndromes have been described. A late chronotype is a risk for circadian misalignment, which impacts the regulation of seizures. Several studies suggested that primary generalized epilepsy subjects are more likely to have a late chronotype phenotype [
]. Data concerning endogenous melatonin secretion indicate that patients with epilepsy tend to have a late circadian phase. Subjective measures of chronotype do not specify an evening-oriented chronotype in epilepsy. However, data concerning endogenous melatonin secretion suggests that these subjects manifest a late circadian phase [
]. Circadian rhythms of patients admitted for long term EEG and video monitoring, used measurement of the dim light melatonin onset (DLMO). The data showed that when correlating seizure timing to the circadian phase as measured by the DLMO, temporal seizures occurs in the 6 h before DLMO and frontal seizures mainly in 6–12 h after the DLMO. Temporal and frontal seizures occur in a non-random fashion synchronized to a hormonal marker of the circadian timing system [
]. The mean Morningness-Eveningness Questionnaire (MEQ) score was reduced in subjects with epilepsy compared to those with focal epilepsy (FE), but was not significantly lower than that in healthy controls. The mean DLMO time in subjects with epilepsy was 49 min later than that in healthy subjects. A lower increase in melatonin within the 30 min after DLMO has been described in epilepsy [
Associations between circadian rhythm-related transcription factors regulating clock genes and the mTOR (mammalian target of rapamycin) signaling pathway have been suggested. Rhythmic activity of hyperactivated mTOR signaling molecules leads to a circadian augmentation in neuronal excitability. Oscillations of neuronal excitability in the SCN have been proposed to modulate the periodic excitability in the hippocampus via the subiculum by long-range projections [
Rhythmic patterns in epileptic activity and seizure occurrence correlate with specific conditions and circadian variation in excitatory and inhibitory equilibrium. The core circadian genes BMAL1 and CLOCK affect excitability and seizure threshold [
]. Period1 (Per1) is a clock-oscillating gene product, which regulates the circadian rhythm in the hypothalamic SCN. Per1 is also expressed in additional brain areas including the cerebral cortex, amygdala, and hippocampus, suggesting it is involved in wider cellular functions besides the control of rhythm. Chemical or electrical seizure-inducing neuromodulation controls Per1 expression. Electric convulsive shock (ECS) and kainic acid (KA) promote the expression of Per1 mRNA in the hippocampus and cortex [
]. In a pre-clinical study of mTLE, the circadian phase and strength of activity were altered in early post-SE and epileptic stages. The acrophase of the spontaneous locomotor activity (SLA) rhythm was delayed throughout epileptogenesis and had a fragmented 24 -h rhythmicity. A protracted active stage length was also documented during the epileptic phase. The expression of Bmal1, Cry1, Cry2, Per1, Per2, and Per3 was altered. The diurnal rhythm of Cry1 and Cry2 was absent in the early post-SE and recovered during the epileptic phase. Per1 and Per2 rhythmic expressions were disturbed in post-SE groups [
]. Attenuated oscillation of several core clock genes was correlated with disturbed diurnal and circadian rest-activity, and sleep-wake patterns, in Kcna1-null mice. EEG confirmed changes in sleep patterns, with more time spent awake and less time spent asleep [
]. Memory deficits in subjects with DRE may be affected by exposure to cortisol. In a study of 52 adults with DRE, those with reduced memory scores had increased cortisol and a slower decline in afternoon levels [
]. A diurnal configuration was described for cortisol and HRV, shown by increased levels of physiological arousal in the mornings and decreased levels at night in subjects with epilepsy and psychogenic non-epileptic seizures (PNES) [
]. Vasopressin function is phase-conserved among diurnal and nocturnal species. A decreased level of melatonin was described in patients with refractory epilepsy and was proposed to be a consequence of the natural course of epilepsy or a result of ASD [
]. Pinealectomy causes convulsions in some species, and melatonin partially protects against this pro-convulsive effect. Melatonin also increases the current required to induce epileptic after-discharges in a rat model of epilepsy [
The potential role of chronobiology in epilepsy was proposed as a basis for the development of chronotherapy-based modalities, which may have some benefit in DRE. Chronopharmacology, chronopharmaceutical delivery methods, differential medication dosing, and utilization of zeitgebers, including chronobiotics or light-therapy and desynchronization, have been suggested [
]. The three features of chronopharmacology are the chronobiology of disease, the pharmacokinetics of the drug and its relation to circadian rhythms, and the interaction between biological rhythms and pharmaceutical action. Personalization to a patient's chronotype can benefit chronopharmacology and chronotherapy. These can be assessed by questionnaires like MEQ or by direct measurements of DLMO, actigraphic factors, cortisol production, and sleep parameters [
In a preclinical study, large circadian variations in VPA pharmacokinetics were described, and were suggested to be associated with the circadian rhythm in toxicity, as the ideal tolerance paralleled the time at which it induced the lowermost values of C(max) and AUC. A single dose of VPA was administered and matched for 3 weeks to a 12 -h light (rest span) and a 12 -h dark cycle (activity span). The AUCn(0-infinity) was higher when VPA was injected at 19 h after light onset (HALO) compared to the administration at 7 HALO. Dosing at 7 HALO induced the maximum Cl(T) value, whereas the Cl(T) was slower when VPA was administered at 19 HALO. C(max) and AUCn(0-infinity) have a significant circadian rhythm [
]. The effect of injection time on tolerance to VPA was studied in mice synchronized under a light-dark cycle (12:12). The best tolerance to VPA was shown when the drug was injected in the second half of the light-rest span, which corresponds to the second half of the night for humans. Following a lethal dose, the surviving mice showed a circadian variation in body temperature and body weight loss, with the least temperature change and weight loss occurring when the VPA was administered at 9 HALO. Lethal toxicity also varied according to circadian dosing-time [
Awakening epilepsy occurs more frequently in the late-night hours while focal motor seizures are more common in the early part of the night. The effects of chronotherapeutic dose schedule of phenytoin and CBZ were studied in a group of patients with blood serum levels at subtherapeutic levels and patients with toxic serum levels. Of the 103 patients in the subtherapeutic level group, 51 patients (STG I) received a constant medication dose and the other 52 patients (STG II) were given the majority of the dose at 20:00. Of the 63 patients in the toxic level group, 31 patients (TG I) received a lower dosage at regular times, while the other 31 patients (TG II) received most of the dose at 20:00. The study showed that patients in the STG II group had therapeutic levels of the drug and improved seizure control than STG I patients, and TG II patients had improved drug tolerance compared with those in the TG I group. The data show that a dosing schedule can improve the effect in diurnally active patients with epilepsy in whom standard treatment regimens failed [
]. Using a clobazam dosing regimen, which is tailored to the timing of patients’ seizures, improves its efficacy in controlling seizures. An increased evening dose regimen as add-on therapy was evaluated in subjects with night-time/early-morning seizures based on an increased rate of seizures (> 80 %) at nighttime. Subjects with differential dosing tolerated an increased median total clobazam dose. The median fraction of the total clobazam dose delivered in the evening was 66 % and was associated with a median seizure reduction of 75 % compared to only 50 % in controls. Overall the data supported that an increased-evening differential dose of clobazam enhanced seizure control in patients with night-time and early-morning seizures, supporting the need for tailoring treatment to individualized chronobiology [
Melatonin is associated with the control of circadian rhythm and exerts a neuroprotective and anti-seizure effect. In a model of TLE, melatonin improved the seizure-latent period, reduced the incidence of spontaneous recurrent seizures (SRSs), and decreased the circadian rhythm of seizures. Melatonin decreased neuronal damage in the hippocampus and piriform cortex [
Treatment with melatonin after status epilepticus attenuates seizure activity and neuronal damage but does not prevent the disturbance in diurnal rhythms and behavioral alterations in spontaneously hypertensive rats in kainate model of temporal lobe epilepsy.
]. Agomelatine is an anti-depressant, which functions as an agonist to melatonin MT1 and MT2 receptors, and as an antagonist to the serotonin 5HT2C receptor. It also reduces the depolarization-evoked release of glutamate, induces a neuroprotective action, and has been proposed as a treatment for epilepsy [
1.5 Variability in biological systems and response to ASD
Variability characterizes the functions of many biological systems that are dynamic and continuously change over time. These systems are mandatory for the responses to internal and external triggers that maintain homeostasis and normal function [
In epilepsy, changes noted in seizure occurrences do not reflect alterations in the risks for the development of seizures. Alterations in seizure incidence occur due to probabilistic variation in seizure risks initiated by normal historical fluctuations, leading to unpredictability in seizure events. No statistical approach can distinguish predictable changes in the incidence of seizure events due to natural variability from changes in the underlying risks for these events. Using a database of 1.2 million recorded seizure events over 8 years, an epilepsy seizure risk assessment tool (EpiSAT), which uses a Bayesian mixed-effects hidden Markov model, was developed. The score was able to predict alterations in the risk for seizure events using a subject-reported seizure diary and clinical information, and differentiate real changes in the risk from natural variations in the frequency of seizures [
]. Chronic EEG studies indicated that epilepsy is a cyclic disease with rhythms operating over many different time scales: multi-day (multidien), circadian, and seasonal. However, seizure events were not evenly spread over time, and were clustered at several periods. The multidien rhythms were not synchronous across animals, suggesting an endogenous generator [
]. Prediction of a seizure using chaos analysis of signals from EEG has been proposed. A chaos attractor was reconstructed in the phase space of EEG. Entropy parameters calculated from the Lyapunov exponents showed that prior to the seizure, these parameters had lower values when compared with healthy conditions [
]. The high degree of variability in many systems is also reflected in the drug response. Both inter- and intra-patient variabilities, which are beyond the expected pharmacodynamics and pharmacokinetics of the medications, have been described [
Personalized inherent randomness of the immune system is manifested by an individualized response to immune triggers and immunomodulatory therapies: a novel platform for designing personalized immunotherapies.
]. Interactions of the drug with its cellular target cannot always be defined by simple diffusion and intrinsic chemical reactions. In many systems, multiple non-specific interactions between macromolecules in cells and medications are involved in determining the effect. These are not considered by simple pharmacodynamics and pharmacokinetic rules. The lack of specificity has been shown to be associated with a slow incorporation in the kinetics of DNA-binding drugs. Non-specific interactions have been demonstrated in several cellular compartments, as shown by a high degree of variability in intracellular drug kinetics [
]. A marked daily variability in drug serum levels has been described for a single subject. This intra-patient variability further supports the notion that this is a complex dynamic system reflected by a high level of unpredictability [
]. Subjects with apparently similar types of epilepsy show different responses to the same ASD. Responsiveness and resistance to ASD were described in a pre-clinical model, along with resistance to multiple drugs [
]. In most pediatric patients with epilepsy treated with levetiracetam, there is no strong indication for a concentration-response association with efficacy or toxicity. It has been suggested that concentrations of levetiracetam do not correlate with the therapeutic value, implying that clinical judgment, without assessing drug concentrations, should be the main strategy [
]. Genetic variability has been observed in the ABCB1 gene coding the efflux transporter multidrug resistance protein 1 (MDR1). However, population-based studies of ABCB1 polymorphisms differ in their conclusions [
]. The intra-subject variation in serum levels of levetiracetam (LEV), topiramate (TPM), and lamotrigine (LTG) following generic switch has been studied. An increase or decrease of a > 20 % alteration in drug concentration in the plasma was reported, while the patients’ received stable dosages, has been reported by therapeutic drug monitoring (TDM) studies. Both inter-day variability in intra-patient LEV, TPM, and LTG plasma levels were described, even in subjects stabilized with the same drug over a long period of time [
Taken together, the data suggest a high degree of variability in seizures and response to ASD, which can also partly explain the high rate of resistance that develops in these patients.
1.6 A stepwise approach for establishing a platform for overcoming drug resistance in epilepsy
Many biological systems are implementing variability patterns in their function as part of a process to identify the optimal response to different triggers. An “optimal state of variability” has been proposed to take on a ‘U shape’ representing chaotic variability in a steady state and complete predictability [
]. Intermittent dosing can recover clinical benefits, while reducing drug exposure and adverse effects. An intermittent dosing regimen for the anti-seizure effect of rapamycin, an mTOR inhibitor, was studied in a tuberous sclerosis complex (TSC) mouse model. Intermittent dosing of rapamycin using drug holidays of > 3 weeks preserves a substantial anti-seizure effect [
]. In a prospective trial of patients with inflammatory bowel disease receiving biological therapy, loss of clinical response was noted in 36 % of patients on regular regimens vs. only 13 % of patients who were on a dose alteration regimen, irrespective of the lack of change in drug serum level [
]. Version 1.0 implements a random-number generator for introducing variability in dosing and time drug administration. This is done within the therapeutic window of the drug, or multiple drugs, that the patient is using. The introduction of “simple” variability is expected to overcome drug-resistance and improve the response to the medications that the patient is currently taking. Version 2.0, implements a closed-loop algorithm where inputs are based on clinically meaningful endpoints for generating therapeutic regimens. For, version 3.0, host and disease-related patterns of variability are quantified in a personalized manner, and then implemented into a true-random number generator. This version involves a process of continuous adaptation of algorithm output to inputs from quantifiable variability parameters. These include parameters associated with disease pathogenesis, host response, and the mechanism of action of the drug [
]. In the second stage, closed-loop machine learning is implemented into the algorithm along with the introduction of several epilepsy-related ANS and chronobiology parameters. The large interdependency between multiple network properties in these systems, many of which behave randomly, is isolated and quantified. These personalized signatures are used as a means for implementing variability patterns in an individualized way. Epilepsy is characterized by periodic dynamics that augment the likelihood of seizures at certain times of the day, and which are vastly patient-specific. Predictive models for epilepsy are improved by circadian information [
]. Deep phenotyping, EEG-based studies of network dysfunction, quantifiable brain imaging, innate immunity processes, microRNA as therapeutic targets, and genetic variants have been proposed as tools for improving personalized therapy in epilepsy and can be implemented into the algorithm [
Fig. 1 presents a schematic presentation of a closed-loop system into which multiple epilepsy parameters are introduced and a personalized tailored treatment is offered to improve the effect of ASD while reducing toxicity.
In summary, DRE is a multifactorial process, affecting up to a third of all subjects with epilepsy. The ANS and chronobiology may contribute to the pathogenesis of the disease. A platform consisting of a closed-loop system, in which parameters based on ANS, chronotherapy, and variability are implemented in a personalized way, is being developed with an aim of improving the response to current medications, while reducing toxicity. Ongoing trials (NCT03843697; NCT03747705) have been designed to assess the implementation of these algorithms in patients and will provide insight into various underlying mechanisms, enabling the generation of improved treatment modalities. Additional studies are required for assessing the role of chronobiology in DRE and the potential for improving the response to anti-seizure medications using chronotherapy.
Declaration of Competing Interest
The authors declare no conflict of interest.
YI is the founder of Oberon Sciences.
Descriptive epidemiology: prevalence, incidence, sociodemographic factors, socioeconomic domains, and quality of life of epilepsy: an update and systematic review.