In 2018, the Food and Drug Administration (FDA) approved the first-in-class drug derived from the cannabis plant. Although the precise mechanism by which the cannabidiol (CBD) exerts its anti-seizure effects is still poorly known, it seems not to act through interaction with known cannabinoid receptors (15), but holds an affinity for multiple targets, resulting in the reduction of neuronal excitability which is relevant for the pathophysiology of the disease (16, 17).

Cannabidiol is approved for the treatment of seizures in children with DS or Lennox-Gastaut syndrome (LGS) aged 2 years or older, based on three pivotal phases 3 trials (12). In 2019 CBD gained approval in Europe in conjunction with clobazam (CLB), based on clinical trial data showing that the combination of both CBD and CLB resulted in greater efficacy outcomes (16).

The first clinical trial (17) included 120 patients with DS aged between 2 and 18 years. The median frequency of convulsive seizures decreased from 12.4 to 5.9 per month, as compared with a decrease from 14.9 to 14.1 per month with the placebo. Furthermore, 43% of patients in the active arm and 27% in the placebo group showed at least a 50% reduction in the convulsive-seizure frequency. Overall, 62% of patients under CBD did gain at least one category at the seven-category Caregiver Global Impression of Change scale, as compared to 34% in the placebo group. Five percent of patients under CBD became seizure-free, while none in the placebo group did.

Another randomized, double-blind, placebo-controlled, trial (18) included 171 LGS patients aged between 2 and 55 years and measured the reduction in drop-seizures. The median percentage reduction was 43.9% in the CBD group and 21.8% in the placebo group. In 2018, Devinsky et al. (19) compared a lower 10 mg/kg/day dose of CBD with the full 20 mg/kg/day in LGS patients. A median 41.9% reduction in drop-seizure frequency was observed in the 20-mg CBD group, while the median reduction was 37.2% in the 10-mg group and 17.2% in the placebo group. Although this study demonstrated patients may gain benefit in seizure reduction by increasing the dose to 20 mg/kg/day, it also displayed an increased risk in AEs. It is generally recommended to begin at 5 mg/kg divided into two intakes a day, then increase to 10 mg/kg/day. If the 10 mg/kg/day dose is well-tolerated and the anti-seizure effect continues, dosing can be increased to the maximum of 20 mg/kg/day (15).

Cannabidiol also proved to effectively reduce seizure frequency at long-term follow-up (20), retaining a consistent reduction (between 42.9 and 44.3%) in seizure frequency at 48 weeks of follow-up. Moreover, 5 out of 104 patients (4.8%) were convulsive seizure-free at 12 weeks of treatment, with more than 40% having a reduction of convulsive seizure frequency ≥50% at each programmed visit of follow-up (18). In terms of median percentage reduction in convulsive seizures, rates of responders, reduction in total seizures, and CGIC-assed improvements, CBD proved greater in the subset of patients concomitantly treated with CLB. Moreover, the combination CBD+CLB showed a benefit in the number of convulsive seizure-free days (16). However, a drug-to-drug interaction increasing levels of active metabolites of both compounds must be assessed and hence CLB dose reduction is recommended if patients experience somnolence or sedation (15, 16).

In conclusion, RCTs settle CBD as a well-tolerated drug, with patients primarily experiencing somnolence, diarrhea, and decreased appetite. The elevation of liver transaminases may be observed mostly in patients on concomitant valproate, and the dose reduction of valproate or CBD is often decisive. The efficacy of CBD on both convulsive and drop seizures is established, with retained efficacy at long-term follow-up. New RCTs in other syndromic or isolated epilepsies populations may widen the field of use of CBD in the next few years.

Fenfluramine (FFA), formerly used at 10 times higher dosage (up to 120 mg/day) as a weight-loss drug, exerts its anti-seizure effect both through the release of serotonin which stimulates multiple 5-HT receptor subtypes, and by acting as a positive modulator of sigma-1 receptors (16, 21–23). Fenfluramine has been approved by the FDA in June 2020 and is currently under evaluation by the European Medicines Agency (EMA). The drug proved significantly effective in reducing seizures in phase-3 trials on DS patients: the 0.8 mg/kg/day treated group did experience a mean 64% reduction in seizures as compared to 34% in the 0.2 mg/kg/day group. Notably, a >75% reduction in seizures occurred in 45% of patients under 0.8 mg/kg/day, in 20.5% of those on 0.2 mg/kg/day compared to 2.5% in the placebo group (23). Fenfluramine has then continued to provide a clinically meaningful reduction in convulsive seizure frequency over a median of 445 days of treatment. The median percent reduction in monthly convulsive seizures frequency was 83.3%. Overall, 62% of patients showed a 50% reduction in convulsive seizure frequency (16).

Together with the anti-seizure effect, FFA has also relatively few drug-drug interactions, primarily a moderate effect on stiripentol (STP), which requires the downward adjustment of FFA dosing to.5 mg/kg/day. No additional interaction with other drugs such as valproate, CLB, and CBD are known (15). The most common AEs reported under FFA treatment include decreases in appetite, weight loss, diarrhea, fatigue, lethargy, and pyrexia (16). The main AEs leading to FFA withdrawal as a weight-loss agent were the occurrence of valvular heart disease (VHD) and pulmonary arterial hypertension (PAH), for which 6-month-echocardiographic monitoring is required together with an ECG. However, at the anti-seizure dosages, no VHD or PAH was observed after a median duration treatment of 256 days. No ECG alterations indicative of atrioventricular conduction or cardiac depolarization alterations were seen, and no mitral or aortic valve regurgitation greater than “trace” was observed in any of the 232 patients with DS who participated in the open-label extension (OLE) study (21, 24, 25).

Cenobamate (Xcopri or YKP3089) is a new ASM that has recently gained approval by the FDA for the treatment of focal-onset seizures in adults. The EMA is currently reviewing the drug for approval as an adjunctive treatment in focal-onset epilepsies. Cenobamate is a tetrazole-derived carbamate compound with a dual mechanism of action; the drug can both enhance the inactivated state of voltage-gated sodium channels, and act as a positive allosteric modulator of the GABA-A receptors, binding at a non-benzodiazepine site (26).

A multicenter, randomized study of patients with uncontrolled focal seizures (27) showed that the adjunctive cenobamate, with dosage groups of 100, 200, and 400 mg/day led to a consistent reduction in focal-seizures frequency after 18-weeks of treatment, with the greatest reduction observed in the 200 and 400 mg/day doses groups. A similar dose-effect relationship was seen when evaluating the responder rates (≥50% in seizure reduction). Post-hoc analysis proved seizure frequencies decreased early during cenobamate titration; while, during the 12-week maintenance phase, significantly more patients under the active 200 or 400 mg/day harms achieved seizure freedom as compared to that receiving placebo. Cenobamate is overall well-tolerated, showing mild to moderate severity AEs on the CNS system, such as somnolence, dizziness, and disturbances in gait and coordination, with a linear incidence-dose correlation and disappearance at maintenance. Four cases of hypersensitivity adverse reactions occurred during two RCTs, including one serious AEs of Drug Rash with Eosinophilia and Systemic Symptoms (DRESS) (26, 27). In this case, the rapid titration of 100 mg/week from 200 to 400 mg dose might have contributed to the higher rates of AEs in the 400 mg group; a lower starting dose and a slower titration rate have been shown to reduce the occurrence of hypersensitivity reactions, possibly through the development of immune tolerance (27). As cenobamate inhibits the P450 family cytochrome CYP2C19^*18, dosing adjustment is needed when adding cenobamate to ASM regimens containing phenytoin or phenobarbital (28); moreover, a dose reduction of CLB should be considered, counteract the increase in plasma levels of desmethylclobazam, its active metabolite. Cenobamate has also been shown to decrease by 25% the plasma exposure to carbamazepine, through the induction of the CYP3A4. Cenobamate could shorten the QT-interval on the ECG in a dose-dependent manner. Hence, cenobamate is contraindicated in patients with familial short QT syndrome, and caution is required in co-administration with other drugs known to reduce the QT interval since a synergistic effect may occur (26, 27). In a short time, data will help to assess cenobamate active time-window on seizures control and real-life data will help to acknowledge whether freedom rates will be borne out in clinical practice. The mechanisms of action and the potential additive or synergistic interactions of cenobamate with concomitant ASMs also warrant further investigation (26).

Vagal nerve stimulation (VNS) is approved both in Europe and in the United States as an adjunctive treatment in patients with refractory epilepsies, and it is routinely available in many epilepsy centers, with more than 100,000 patients treated worldwide (6). Vagal stimulation may then turn off seizures originating in regions susceptible to heightened excitability, such as the limbic system, thalamus, and thalamocortical projections. Moreover, an additional mechanism of action derives from the activation of the locus coeruleus and the raphe nuclei, and the regulation of the downstream release of norepinephrine and serotonin, both having antiepileptic effects (29).

Two large RCTs showed VNS efficacy in reducing seizures, achieving a 50% reduction in 31% of patients, and over 50% seizures reduction in 23.4% of the studied population. On the other hand, seizure freedom at long-term follow-up was observed in <10% of patients. Side effects are usually mild and include hoarseness, throat paraesthesia or pain, coughing, and dyspnea. This tends to improve over time or through the adjustment of setting parameters (6).

In conclusion, evidence suggests VNS is well-tolerated in both children and adults with drug-resistant partial epilepsies (30–32); moreover, the newest VNS models can detect ictal tachycardia and automatically deliver additional stimulation to abort seizures or reduce their severity (6).

Developed as a non-invasive alternative to VNS, the transcutaneous VNS (tVNS) acts on the auricular branch of the vagus nerve (ABVN), targeting thick-myelinated afferent fibers in the cymba conchae, and hence activating the ipsilateral nucleus of the solitary tract (NTS) and locus coeruleus. This activation pathway overlaps with the classical central vagal projections, leading to a brain activation pattern similar to that produced by invasive VNS (33). The device consists of a programmable stimulation apparatus and an ear electrode (34). Stimulation setup is adjusted by applying decreasing and increasing intensity ramps and achieving a level just above the individual detection threshold, but clearly below that of pain. Patients usually apply tVNS for 1 h/three times per day (33) and adherence is usually high (up to 88%) (35). Trials converge in demonstrating up to 55% reduction in seizure frequency, with mild or moderate side effects mainly including local skin irritation, headache, fatigue, and nausea (6, 35).

Deep brain stimulation (DBS) is a minimally invasive neurosurgical technique, which through implanted electrodes can deliver electrical stimuli to deep brain structures. Patients with refractory focal epilepsies and not eligible for surgery are usually good candidates (29). Both stimulation of the ictal onset zone and the anterior thalamus have gained approval by the FDA as effective stimulation sites, providing a significant and sustained reduction in seizures together with the improvement of the QoL. Nowadays, both DBS and responsive neurostimulation (RNS) are available, being the latter a system able to monitor electrical changes in cortical activity and give small pulses or bursts of stimulation to the brain to interrupt a seizure (36). The interim results of a prospective, open-label, and long-term study did show that the median 60% or greater reduction in seizure frequency is retained over years of follow-up. Moreover, the majority of patients took advantage of treatment with the RNS® System, and 23% experienced at least one 6-month period of seizure freedom (37). The most relevant reported side effects were depressive mood and memory impairment, besides the local side effect of implantation. Nonetheless, it should be stated that RNS is a feasible option in most epilepsy centers in the US, but its use remains limited in other parts of the world. In these cases, DBS could be an option with targets and stimulation parameters selection are largely driven by the experience of the referred center (38, 39).

Trigeminal nerve stimulation (TNS) is a novel neuromodulation therapy, designed to deliver high frequencies stimulation in a non-invasive way, hence modulating mood and relieving symptoms in drug-resistant epilepsies. The study of DeGeorgio et al. (40) found that the responder rate (at least 50% reduction in seizures) was 30.2% in the active group, while it was 21.1% in the control group. Moreover, the responder rate did further increase over the 18-week treatment period in the actively treated group. TNS was overall well-tolerated and, when occurring, treatment-related AEs were mild including anxiety (4%), headache (4%), and skin irritation (14%). However, long-term follow-up studies showed inconclusive results (6), meaning further studies and patient monitoring will be needed in the next years.

The transcranial direct current stimulation (tDCS) displays the use of two skull electrodes (anode and cathode) to induce widespread changes of cortical excitability through weak and constant electrical currents. Cortical excitability may increase following anodal stimulation, while it generally decreases after cathodal stimulation. Based on this principle, hyperpolarization using cathodal tDCS has been proposed to suppress epileptiform discharges. Major six clinical studies are promising with 4 (67%) showing an effective decrease in epileptic seizures and 5 (83%) exhibiting a reduction of epileptiform activity. However, some results may be misleading due both to the small and heterogeneous nature of the studied populations and to the different setting parameters applied. Hence, nowadays the major achievement is the demonstration that tDCS may be effective and safe in humans; however, further studies will be needed to define setting stimulation protocols and understand the long-term tDCS effectiveness (41).

The nerve cells of a brain to a maximum depth of 2 cm can be stimulated using transcranial magnetic stimulation (TMS). To this, low-frequency and repetitive magnetic stimulations have been shown to induce long-lasting reductions in cortical excitability and, hence, have been proposed as a treatment for drug-resistant epilepsies (4). Probably, it is the repeated nature of magnetic pulses which allows modulating the neuronal activity, wherein high frequencies (>5 Hz) would have an overall excitatory effect, while low-frequencies (0.5 Hz) would exert an inhibitory effect on neurons (29).

Despite the optimal stimulation parameters still needing to be clearly defined, they are likely of crucial importance because treatment intensity depends both on the number of pulses and the number of sessions applied over the treatment period. Superior results are achieved in patients with neocortical epilepsy, whit a calculated effect size of 0.71 and 58–80%. This makes sense taking into account the rapid decay of the strength of the magnetic field with distance hence no adequate secondary currents can be elicited in the deep cortex. However, evidence suggests the effects of repetitive TMS may not be restricted to the only site of stimulation but may spread from focal areas to wider areas of the brain.

In conclusion, results should be reproduced in larger cohorts with double-blinded randomized trials, but are promising if compared to the effects currently achieved with invasive neurostimulation techniques for the treatment of epilepsy (42).

The Interleukin IL-1R1-TLR4 signaling pathway originates the neuroinflammatory cascade in epilepsy through increased levels of either the endogenous agonists or their receptors, or even a combination of both (2). These findings prompted the clinical use of anakinra, the recombinant, and modified form of the human IL-1Ra protein. Case report studies of Anakinra in patients with intractable seizures did result in a significant reduction of seizure activity and improvement of cognitive skills (44). Moreover, IL-1R1 and TLR4 signaling have been targeted by specific, non-viral, small interfering RNAs (siRNAs) to knock down the inflammasomes or caspase 1 in rats with kindling-induced SE (45).

Prostanoids are a family of lipid mediators generated from the cell membrane arachidonic acid by cyclooxygenase enzymes 1 and 2 (COX1 and COX2). Prostanoids bind to specific G protein-coupled receptors (GPCR), hence regulating both innate and adaptive immunity (46).

Activation of the Toll-like receptors (TLRs) can lead to reactive oxygen species (ROS) production, hence promoting and sustaining inflammatory pathways. The detrimental effects of ROS are usually counteracted through the activation of the nuclear factor E2-related factor 2 (Nrf2). Activated Nrf2 translocates to the nucleus where it heterodimerizes with the small Maf proteins (sMaf) and binds to the antioxidant response element (ARE 5′-TGACXXXGC-3′) battery activating transcription of genes that are involved in antioxidant and cytoprotective tasks (53).

Transient administration of N-acetyl-cysteine (NAC), a glutathione precursor, did prove to activate Nrf2 in mouse models of SE, thus inhibiting high mobility group box 1 (HMGB1) cytoplasmic translocation in the hippocampal neural and glial cells and preventing the linkage between oxidative stress and neuroinflammation for which the redox-sensitive protein HMGB1 is central (2). Also, high doses (4–6 g/day) of NAC were used in Unverricht-Lundborg disease (ULD), progressive myoclonus epilepsy (PME), showing overall improvement of myoclonus, ataxia, and generalized tonic–clonic and absence seizures. Neuroprotection and improvements in spatial learning abilities were also observed with retained beneficial effects during treatment (54, 55).

Adeno-associated viral (AAV) vectors gene delivery may provide long-term, persistent, induction of Nrf2 expression in a variety of cell types in the brain, with minimal toxicity. The injection of AAV coding for human Nrf2 in the hippocampus of mice with spontaneously recurrent seizures resulted in a reduction in the number and duration of generalized seizures, which interestingly was performed in the already established epileptic phase, highlighting the direct potential of such interventions in the treatment of epilepsy (56).

Viral gene therapy may employ three classes of viral vectors, namely, adenovirus (AD), adeno-associated virus (AAV), and lentivirus. All these three viral vectors have successfully demonstrated to attain high levels of transgene delivery in in-vivo disease models and clinical trials. However, the risks of immunogenic responses and transgene mis-insertions, together with problems in large-scale production are still a deal to face (70).

Adeno-associated viruses belong to the Parvoviridae family and proved to retain favorable biology, leading their recombinant forms (rAAVs) to become the main platform for current in-vivo gene therapies (29). A limited clinical trial on patients with late-infantile neuronal ceroid lipofuscinosis (LINCL) did prove neurosurgical gene therapy to be practical and safe, supporting the potentialities of this kind of approach (71). However, in the view of removing invasiveness, interest was moved to engineered capsid which can confer the ability to cross the BBB and transduce astrocytes and neurons, allowing direct intravenous injection. This was achieved through a process of directed selection in a mouse strain, and further work would be needed to develop a similar variant for use in humans (6, 72).

Retroviruses such as lentivirus share with AAVs the ability to infect neurons and lead to a stable expression of transgenes. Lentiviral vectors (lentivectors) are RNA viruses and the transgenes can integrate into the host genome through the reverse transcriptase gene. However, possible insertional mutagenesis may be reduced by using integration-deficient lentivectors, which simultaneously ensure stable transduction (73). Lastly, lentivectors can package larger genes or regulatory elements as compared to AAVs (6).

Different viral vectors intrinsically tend to infect different neuronal and glial subtypes, but the high specificity of the target is far from their properties. Hence, several efforts have been made that to identify specific neuron-type targeting promoters: the calcium/calmodulin-dependent protein kinase II (CamKII) promoter is suitable to manipulate excitatory neurons in the forebrain; on the other hand, targeting inhibitory interneurons may be difficult as promoters for specific GABAergic neurons are poorly defined (6). Finally, the optimal promoter should provide the expression of a level of transgene which is sufficient to moderately alter cell properties but avoids cytotoxicity (6, 74).

As for the transgene, gene therapies have been commonly built on the basis that the excitation–inhibition balance is altered in epilepsy. Hence, on a general principle, gene therapy may work through modulating the expression of neuropeptides, and regulation of the neuropeptide Y (NPY) did already show promise, acting both on pro-excitatory Y1 and pro-inhibiting Y2 receptors in the hippocampus (6, 75). Another way may be that of regulating potassium channels; overexpression of the potassium channel Kv1.1 proved effective in preventing epileptogenesis in a mouse model of focal epilepsy, the physiological basis may lie on the modulation of both neuronal excitability and neurotransmitter release (76, 77). Lastly, chemogenetics refers to the possibility to use gene transfer to express receptors that are insensitive to endogenous neurotransmitters but highly sensitive to exogenous drugs, in a receptor-to-drug therapeutic approach. This promising approach will also allow adjusting the activating drugs to find the optimum dosage with low interference with normal brain function but efficiently suppressing seizures (6). Further refinements of chemogenetics have jet got underway, which may use receptors detecting out-of-range extracellular elevations of the concentration of glutamate and, therefore, inhibiting neurons, preventing drug administration. Although attractive, this strategy will need further work to assess the risk of immunogenicity (6).

Some of the issues of viral vector-based gene therapy may be overcome by non-viral gene strategies, which provide advantages with regards to the safety profile, localized gene expression, and cost-effective manufacturing. Non-viral gene delivery systems are engineered complexes or NPs composed of the required nucleic acid (pDNA or RNAs) and other materials, such as cationic lipids, peptides, polysaccharides, and so on (70). These vectors have low production costs, can be topically administered, can carry large therapeutic genes, use expression vectors (such as plasmids) that are non-integrating, and do not elicit detectable immune response also after repeated administrations (29, 70). Cationic lipid-based vectors are currently the most widely used non-viral gene carriers. Limitations may include low efficacy due to the poor stability and rapid clearance, or the possible generation of inflammatory or anti-inflammatory responses. Hence, cationic polymers, such as poly(L-lysine) (PLL) or modified variants (PEGylated PLL), constitute alternative non-viral DNA vectors that are attractive for their immense chemical diversity and their potential for functionalization (69).

Oligonucleotides are unmodified or chemically modified single-stranded DNA sequences (of up to 25 nucleotides) that hybridize to specific complementary mRNAs. Once bound to targeted mRNAs, oligonucleotides can either promote RNA degradation or prevent the translational machinery through an occupancy-only mechanism, referred to as steric blockage. Anyhow, the process leading to protein formation is inhibited. Synthesizing antisense oligonucleotides (ASOs) must deal with making a structure that must be suitable for a stable and selective oligonucleotide/mRNA complex. Moreover, oligonucleotides are rapidly degraded by endo- and exonucleases and the mononucleotides products may be cytotoxic (29, 78). Hence, the use of ASOs in clinical practice requires overcoming problems related to the design, bioavailability, and targeted delivery (78). To date, few in-human studies have been conducted that primarily addressed invariably progressive and fatal diseases such as PMEs (79, 80). The authors proved the feasibility of the ASOs-based approach by specifically customizing oligonucleotides over the genetic defect of patients. This opens the way to N-of-1 trials, which will hopefully be the road of the next few years not only in oncology but also in epileptic patients (81).