Keywords

14.1 Refractory Epilepsy: Current Explanations

According to the current definition from the International League Against Epilepsy (ILAE) the term refractory (or intractable, or drug resistant) epilepsy refers to the failure of adequate trials of two tolerated, appropriately chosen antiepileptic drug schedules (either as monotherapies or in combination) to achieve sustained seizure freedom (Kwan et al. 2010). In the previous definition “appropriate” indicates an intervention that has previously been shown to be effective (preferably in randomized controlled studies) for the patient’s epilepsy and seizure type, while “adequate” indicates that the drug has been administered at adequate dosage for a sufficient length of time. Regarding what constitutes and adequate period without seizures for a patient to be regarded as “seizure-free,” a minimum of three times the longest pre-­intervention inter-seizure period or 12 months (whichever is longer) has been proposed. Application of a standardized definition of refractoriness is not trivial, since depending on the definition chosen the frequency of drug resistant epilepsy varies considerably (between 10 and nearly 40 %) (Beleza 2009). What is more, as discussed later, general applicability of a given hypothesis of drug resistant epilepsy may critically depend on what we actually call “drug resistant epilepsy.”

Despite the fact that there are currently more than 20 available antiepileptic drugs (AEDs) and that 15 third generation agents have been introduced to the market since 1990, the clinical need of refractory epilepsy remains unmet (Bialer 2012; Löscher and Schmidt 2011): there are still no solid evidence indicating improved efficacy. There are currently four hypotheses explaining the nature of refractory epilepsy: on the one hand, the traditional transporter and target hypothesis (Löscher and Potschka 2005; Schmidt and Löscher 2005; Kwan and Brodie 2005; Remy and Beck 2006); more recently, the inherent severity hypothesis and the neural network hypothesis have also been proposed (Rogawski and Johnson 2008; Fang et al. 2011). Among them, the transporter hypothesis is so far, without a shadow of doubt, the most extensively studied.

The transporter hypothesis suggests that intractable epilepsy may have a pharmacokinetic basis. It states that drug resistance may emerge, as in other disorders, from intrinsic or acquired activation or over-expression of drug transporters involved in drug distribution, metabolism and elimination. Research supporting this hypothesis has focused in efflux transporters from the ATP-binding cassette (ABC) superfamily. Evidence abounds indicating high expression levels of members of this family such as P-glycoprotein (Pgp), breast cancer resistance protein (BCRP), and multidrug resistance proteins (MRPs) at the neurovascular unit of nonresponsive patients (either at the blood–brain barrier or glial cells or neurons) (Tishler et al. 1995; Dombrowski et al 2001; Sisodiya et al. 2002, 2006; Aronica et al. 2003, 2005; Lazarowski et al. 2004; Calatozzollo et al. 2005; Kubota et al. 2006; Ak et al. 2007). Lack of efficacy of those AEDs which are substrates of any of the up-regulated efflux transporter would be a consequence of limited bioavailability of the therapeutic agent in the brain or specifically at the epileptic focus. In fact, some studies showed reduced AEDs concentrations in the brain extracellular fluid and epileptic tissue of refractory patients (Marchi et al. 2005; Rambeck et al. 2006). A general pharmacokinetic mechanism underlying refractory epilepsy is consistent with the fact that available AEDs act through a wide range of molecular mechanisms. The transporter hypothesis has been fully verified in animal models of epilepsy. Several animal models of epilepsy (chronic models) have provided evidence of Pgp over-­expression in brain tissue from animals with refractory epilepsy (Zhang et al. 2012), and drug resistance has been reverted by co-administration of Pgp inhibitors together with the AED (van Vliet et al. 2006; Brandt et al. 2006). Nevertheless, conclusive evidence of the validity of the transporter hypothesis in humans remains elusive. There are some (anecdotical) cases of patients who have showed improvement when AED were co-administered with Verapamil, a known Pgp-inhibitor (Summers et al. 2004; Ianetti et al. 2005; Schmitt et al. 2010; Pirker and Baumgartner 2011). It is still not clear, however, if the observed results are due to the intrinsic antiepileptic activity of verapamil, to Pgp inhibition or another effect on AEDs pharmacokinetics, and randomized control trials with more selective inhibitors are needed to obtain definitive proof of concept. The main argument against the transporter hypothesis is the fact that numerous but not all AEDs are substrates of human Pgp (Zhang et al. 2012). At this point one should bear in mind that current definition of drug resistant epilepsy requires only two adequate, appropriate, well-tolerated AED interventions to consider that a patient presents refractory epilepsy. It is then conceivable that (if the transporter hypothesis were valid) a patient would be diagnosed as drug resistant if at least one of those two AEDs interventions does not include a Pgp-non-substrate (e.g., Carbamazepine). It has been suggested that the transporter hypothesis may be valid for a subgroup of the epileptic patients (Löscher and Delanty 2009).

The target hypothesis states that structural (transcriptional or posttranscriptional) alterations in AEDs molecular targets might explain pharmacoresistance. This hypothesis is based, essentially, in reported loss of sensitivity to voltage-gated sodium channel blockers such as carbamazepine and phenytoin in patients and animal models of epilepsy (Schmidt and Löscher 2009). It has been observed that the inactivation effect of Phenytoin on sodium channels is transiently reduced in kindling models (Vreugdenhil and Wadman 1999), while the use-dependent effect of Carbamazepine and Phenytoin is permanently lost or reduced in the pilocarpine model and in temporal lobe epilepsy patients (Remy et al. 2003a, b; Jandová et al. 2006). Numerous changes in the expression of sodium channels subunits have been described in animal models of seizure and epilepsy, and in epileptic patients (Bartolomei et al. 1997; Gastaldi et al. 1998; Aronica et al. 2001; Whitaker et al. 2001; Ellerkmann et al. 2003), suggesting seizures or epileptogenesis may alter AEDs targets. Mutations at accessory subunit β1 have been linked to a dramatic loss in the use-dependent effect of phenytoin (Lucas et al. 2005). On the other hand, associations between alterations at GABAA receptor subunits and resistance to phenobarbital in animal models of temporal lobe epilepsy have been reported (Volk et al. 2006; Bethmann et al. 2008). The main objection to the target hypothesis is that, as has been already mentioned, there exist clinical AEDs associated to different mechanisms of action. Even those AEDs that share a common mechanism (e.g., GABAA receptor allosteric modulators) frequently bind to different sites of the same receptor. Thus, the target hypothesis by itself would only satisfactorily explain the phenomenon of multidrug resistance involving drugs that share their mechanism of action.

A third hypothesis, the hypothesis of the intrinsic severity, proposes the inherent severity of the disorder as determinant of the treatment outcome (Rogawski and Johnson 2008). It relies on epidemiologic data which indicates that the single most important factor linked to the prognosis of epilepsy is the number of episodes at the early phase of the disorder (MacDonald et al. 2000; Williamson et al. 2006; Sillampää and Schmidt 2006; Mohanraj and Brodie 2006; Kim et al. 2006; Hitiris et al. 2007; Sillampää and Schmidt 2009). Some limitations of the intrinsic severity hypothesis have been highlighted (Schmidt and Löscher 2009): the lack of studies on the biological basis of disease severity; the lack of genetic studies comparing patients with low seizure frequency versus patients with high seizure frequency at the disorder onset and; the fact that there are reports of nonresponsive patients with low frequency of episodes at the early phase of epilepsy (Spooner et al. 2006).

Very recently, a fourth hypothesis has arisen. The neural network hypothesis states that the adaptive remodeling of neural circuits that follows seizures may contribute to the development of refractory epilepsy. However, one should remember that remodeling of neural circuits also occurs in responsive patients. Therefore, differences between the degree of neural reorganization in responsive and nonresponsive patients should be studied to support this latest explanation to drug resistance.

This short overview suggests that either different hypothesis may explain the drug resistance phenomenon in different subgroups of patients (understanding that refractory epilepsy is a complex, multi-factor phenomenon and conceiving that in some patients more than one factor may be present simultaneously) or that the previous hypothesis may be integrated (Schmidt and Löscher 2009), with the two first hypothesis (partially) providing a biological basis for the others. Most importantly to the scope of this chapter, different hypothesis claim for different strategies to develop novel therapeutic answers. In the next sections we discuss potential implications of the first three hypothesis in the field of AEDs development.

14.2 Possible Therapeutic Answers to the Transporter Hypothesis

The obvious answer to overcome efflux transporter-mediated drug resistance is to develop therapeutic systems to circumvent this barrier to achieving adequate concentrations of the drug in its site of action. An excellent review on this matter has recently been published (Potschka 2012). The general strategies studied in the last 15 years to overcome ABC transporters can be synthesized as (Talevi and Bruno-­Blanch 2012): (a) modulation of ABC transporters (i.e., reversal of multidrug resistance and down-regulation of transporters); (b) design of novel drugs which are not efflux transporter-substrates; (c) bypassing drug transport (or the Trojan horse strategy). Most of the research on these strategies has focused on the best known representative of the ABC superfamily, Pgp (note that Pgp was purified back in 1979 and it was not until 1990s that MRPs were identified). However, it is now established that there exist numerous transporters involved in transport of endogenous and exogenous compounds and that the levels of expression of different ABC transporters are interrelated (in some cases, a co-expression pattern has been observed; in others, an inverse relationship has been established) (Miller et al. 2008; Cisternino et al. 2004; Bark et al. 2008; Choi et al. 1999; Bordow et al. 1994). Taking into consideration that the spectra of substrates of different ABC transporters overlap to a certain degree, it might be hypothesized that up-regulation of a given transporter might have a compensatory role in the transient or permanent disturbance of other, which might explain the observed development of tolerance to some interventions aimed at regulating Pgp function (van Vliet et al. 2006). One must consider that development of tolerance is not acceptable when dealing with long-term drug treatments such as AEDs.

Regarding transporters modulation, the most advanced research relates to add-on therapies of specific inhibitors of ABC transporters, a strategy that was originally conceived for cancer treatment. Although preclinical and initial clinical results in the field of cancer treatment were encouraging at first, trials of first, second and even third generation agents had to be stopped at clinical stage due to serious adverse effects (Deeken and Löscher 2007; Lhommé et al. 2008; Tiwari et al. 2011; Fox and Bates 2007). These results have called into question the general validity of this approach of overcoming cellular drug resistance by the use of transporters inhibitors, even though trials continue in order to find more effective and safe inhibitors for Pgp and other transporters (Deeken and Löscher 2007; Akhtar et al. 2011). At this point it is important to remember that ABC transporters comprise a concerted, complex efflux and influx dynamic system whose substrates are not only drugs but also endogenous compounds (e.g., waste products) and toxins. They are implicated in the inflammatory response to several stress and harmful stimuli, and, apparently, they have a role in neurodegenerative diseases such as Alzheimer’s and Parkinson’s disease (Hartz and Bauer 2010). Thus, their permanent impairment or disruption is likely to result in severe side effects (again, one should bear in mind the chronic nature of epilepsy, which demands long-term treatment). A similar outcome to RNA interference technologies to down-regulate a given gene codifying a member of the ABC superfmailiy may be expected (Potschka 2012). Recent research has then focused on elucidating intracellular signaling pathways that control ABC transporters (their expression, intracellular trafficking, activation and inactivation). It is proposed that finding the molecular switches of these transporters will allow selective modulation of transporters function and or expression for therapeutic purposes in different clinical scenarios (Hartz and Bauer 2010), which includes turning the efflux mechanisms off for short, controlled periods of time.

Another strategy which should provide delivery of a drug to the brain without the toxic issues associated to the impairment of the efflux transport is virtual screening or computer-aided design of novel AEDs which are not recognized by ABC transporters (Demel et al. 2008, 2009). A review on in silico models for early detection of Pgp substrates has been recently published (Chen et al. 2012). A 2D QSAR model to detect anti-maximal electroshock seizures (MES) drug candidates (Talevi et al. 2007a, b, 2012), an ensemble of 2D models to identify Pgp-susbtrates (Di Ianni et al. 2011) and a structure-based approach based on homology modeling of human Pgp were jointly applied in a virtual screening campaign to ZINC and DrugBank databases (Irwin and Shoichet 2005; Knox et al. 2011). From 360 compounds predicted as Pgp-non-substrates anticonvulsants, ten diverse candidates (Fig. 14.1) were acquired and tested in the MES test, with good results (Di Ianni et al. 2012, submitted).

Fig. 14.1
figure 00141

A series of novel anticonvulsants emerging from a multistep virtual screening campaign aiming at novel treatments for refractory epilepsy

Full size image

The last strategy implies the application of a carrier system to “hide” the drug from the efflux pump. Different carrier systems have been tested to increase the bioavailability of drugs to the brain, among them nanosystems (polymer nanoparticles, nanogels, lipid nanocapsules, liposomes) (Bansal et al. 2009; Patel et al. 2009; Bennewitz and Saltzman 2009; Alam et al. 2010,). An exhaustive review of all the carriers that have been tested in brain drug targeting to avoid recognition by transporters will deserve an entire chapter or even a book, so we are including some examples for illustrative purposes. A 60-fold increase in the brain localization of doxorubicin (a known Pgp-substrate) in rats, when administered i.v. as polysorbate 80-coated nanoparticles (compared to: i.v. administration in saline solution; in polysorbate 80 solution and; bound to nanoparticles without polysorbate 80 coating) (Gulyaey et al. 1999). Much more recently, a 2.6-fold increase in coumarin-6 localization in the brain through encapsulation of the drug in poly(ε-caprolactone)-block-poly(ethyl ethylene phosphate) nanomicelles was achieved (Zhang et al. 2010). Regarding specific application of this strategy to antiepileptic agents, different nanosystems have been studied for the delivery of Clonazepam, Diazepam, Phenytoin, Ethosuximide, 5-5-diphenyl hydantoin, carbamazepine, and valproic acid (VPA) and NMDA receptor antagonists (Fresta et al. 1996; Kim et al. 1997; Jeong et al. 1998; Nah et al. 1998; Ryu et al. 2000; Darius et al. 2000; Friese et al. 2000; Thakur and Gupta 2006; Abdelbary and Fahmy 2009; Varshosaz et al. 2010; Eskandari et al. 2011). A valid question would be whether this galenic artifices do improve availability of the drug in the central nervous system (CNS) and, if so, the molecular basis of such improvement. Unfortunately, most of this reports limit to physical characterization and in vitro behavior of the proposed systems. However, some of them explore the in vivo behavior of the nanosystems, with variable results. Darius et al. (2000) found that the brain tissue levels of VPA were not altered by administration with nanoparticles, though the nanosystem inhibits metabolism of VPA via mitochondrial beta-oxidation. Friese et al. (2000) reported that poly(butylcyanoacrylate) nanoparticles coated with polysorbate 80 prolong the duration of the anticonvulsive activity of NMDA receptor antagonist MRZ 2/576, presumably by prevention of active transport processes at the choroid plexus. More recently, Eskandari et al. (2011) have found an increased protective effect of VPA in the MES test when the drug was administered in nanostructured lipid carriers to rats. Intranasal administration of a dose of 4 mg/kg of nanostructured lipid carriers of VPA lead to almost three times higher brain concentrations than an intranasally administered solution of 30 mg/kg of the drug; brain–plasma ratio was also increased with the nanosystem (Table 14.1).

Table 14.1 Increment of brain bioavailability of VPA when administered intranasally with nanostructured lipid carriers [adapted from Eskandari et al. (2011)]
Full size table

Prodrugs are another option to circumvent the blood–brain barrier, sometimes making use of influx transporters (e.g., dopamine is administered as its precursor l-dopa, which is transported into the brain by the l-type amino acid transporter and bio-transformed to dopamine in situ) (Mandaya et al. 2010). Numerous prodrugs of different anticonvulsant agents such as phenytoin, gabapentin, VPA and eslicarbazepine have been developed in order to improve bioavailability by regulation of drug absorption, distribution and elimination (Bennewitz and Saltzman 2009; Trojnar et al. 2004; Bialer and Soares-da-Silva 2012). DP-VPA (Fig. 13.2) was designed to be specifically activated at the epileptic focus. It is a prodrug of VPA in which the VPA moiety is covalently bound to a phospholipid, lecithin, leading to a 50-fold increase in efficacy in the pentylenetetrazol-induced seizures test (Trojnar et al. 2004). Similarly, our group has developed prodrugs of VPA with myo-inositol (Fig. 14.2) aiming at capitalizing the active influx of inositol enantiomers into the brain; the activity of these prodrugs in animal models of seizure is also increased compared to VPA, seemingly by improving CNS bioavailability (Bodor et al. 2000; Moon et al. 2007; Bruno-Blanch and Moon 2010). Whether these prodrugs interact with efflux transporters and bypass up-regulated transporter molecules at the neurovascular unit has yet to be studied.

Fig. 14.2
figure 00142

Two prodrugs of VPA designed for improvement of VPA bioavailability at the epileptic focus: a prodrug of myo-inositol (left) and DP-VPA (right)

Full size image

It is noteworthy that in the last few years it has been proven that, besides helping bypassing Pgp, many pharmaceutical excipients which are usually incorporated into carrier-systems can inhibit or modulate Pgp function by different mechanisms (Bansal et al. 2009). For example, it has been proposed that PEG and surfactans such as sorbitans and polysorbates can disrupt the lipid arrangement of the cellular membrane and that these perturbations have been shown to modulate Pgp activity (Lo 2003). This kind of modulation is interesting since it may increase drug bioavailability in a transient manner, without the undesired effects of direct inhibition. Besides its possible role modulating transporters, cumulative evidence indicates that nanoparticle’s coating leads to adsorption of elements from the blood such as apolipoproteins, which in turn allows distribution to the brain by receptor-mediated transcytosis (Wohlfart et al. 2012 and references therein).

14.3 Possible Therapeutic Answers to the Target Hypothesis

Several CNS disorders (either neurological or affective) present a complex etiology which includes a combination of polygenic, environmental, and neuro-­developmental factors. Empiric evidence with effective treatments for some of such diseases (e.g., antidepressants) shows that searching for polyspecific, selective non-selective drugs (multi-target directed-ligands or “magic shotguns” or polyvalent drugs) may prove more safe and effective than the development of highly selective, single-target drugs (Roth et al. 2004). There are plenty examples of recent developments in the field of CNS medications based on this new paradigm, including developing drugs for Alzheimer and Parkinson’s diseases (Cavalli et al. 2008; Youdim and Buccadfasco 2005), schizophrenia, depression and other mood disorders (Decker and Lehmann 2007; Wong et al. 2010).

There are many reasons why multi-target therapies are attractive in the field of epilepsy. First, evidence indicate that—if total drug load is carefully watched—some refractory patients may achieve seizure remission on poly-pharmacy, especially if the pharmacologic properties of the specific AEDs being combined is taken into account (Canevini et al. 2010; Kwan and Brodie 2006). A recent study on 131 patients who underwent successful epilepsy surgery seems to indicate that, at least in the early postoperative stage, dual-therapy may be more effective than mono-­therapy to achieve seizure remission (Zeng et al. 2012). Second, the normal function of neural networks may be more likely preserved by multiple small adjustments than by a single, strong perturbation, reducing not only the likelihood of central side-effects but also the induction of counter-regulatory processes which may relate with drug resistance (Löscher and Schmidt 2011; Bianchi et al. 2009). What is more: many currently used AEDs are in fact unintended multi-target agents (Bianchi et al. 2009).

In the light of the evidence that refractoriness may be in some cases related to modifications in drug targets, the design of novel multi-target AEDs seems as a natural answer to the second hypothesis of drug resistance, considering that it seems to be less likely that two distinct drug targets are altered simultaneously. Therefore, even if one target of a multi-target drug has lost sensitivity, one can speculate that the other/s will remain sensitive.

From the drug design perspective, in silico, rational approaches to develop multifunctional agents can be classified in two strategies (Ma et al. 2010). On the one hand, the combinatorial approach, in which parallel Virtual Screening searches against each target of interest are conducted, retaining those hits that simultaneously gather all the structural requisites needed to interact with each individual target. In other words, the common hits from parallel Virtual Screening searches (one for every model associated to a particular target) are retained. In the background of multi-target drug discovery, the Virtual Screening for ligands for each individual target must be highly sensitive (i.e., a reduced number of false negatives should be observed) since the collective retrieval rate for multiple targets will tend to be relatively low than when aiming to individual targets (one might speculate that, naturally, it is more difficult to find compounds that selectively interact with different targets without being excessively promiscuous). In contrast, when drugs that selectively interact with a single target are being searched, in certain contexts one might sacrifice sensitivity in order to gain specificity. The second strategy is the fragment-based approach. Here, multiple elements or scaffolds that bind to each of the targeted targets are combined (usually through a linker) into a single, often larger molecule. The main drawback of this later approach relates to the poorer pharmacokinetic and toxicological profile of the final drug. Unless small, highly specific blocks/fragments are combined, it is unlikely that a given compound will gather the features for a CNS drug-like drug.

14.4 Possible Therapeutic Answers to the Intrinsic Severity Hypothesis

If the intrinsic severity hypothesis was valid, AEDs research would face to elemental questions. Firstly, what are the determinants of epilepsy severity? And, if the answer to that initial question was answered, how could one control, through a therapeutic intervention, such determinants? During the last 10 years, basic research has begun to provide us some knowledge to attempt some very draft answers to these issues.

Acquired epilepsy is typically initiated by a brain insult followed by a latent, silent period whereby molecular, biochemical and cellular alterations occur in the brain and eventually lead to chronic epilepsy (Waldbaum and Patel 2010a). In the last 10–15 years a link between epileptogenesis and oxidative stress, mitochondrial impairment and inflammation has been established by a large body of studies (Waldbaum and Patel 2010b; Waldbaum et al. 2010; Devi et al 2008; Liang and Patel 2006; Shin et al. 2008; Patel 2004; Sudha et al 2001; Vezzani and Granata 2005; Vezzani et al. 2011; Choi and Koh 2008). These phenomena seem to be both cause and consequence of seizures, constituting a vicious circle which results in a chronic disorder, e.g., inflammatory mediators are released during seizures, and inflammatory mediators take part in seizure generation and exacerbation. It is also interesting to note that chronic inflammation and oxidative unbalance take part in the physiopathology of a diversity of neurological disorders. The brain combines a peculiar set of factors which makes it particularly vulnerable to reactive species: high rate of oxidative metabolism, low antioxidant defenses and abundant polyunsaturated lipids (Devi et al 2008).

In line with the integrative approach towards explaining refractory epilepsy, a series of studies developed in the last decade agree that pro-inflammatory signals and Reactive Oxygen Species play a role in the regulation of ABC transporters’ expression and activity. For example, exposing isolated rat brain capillaries to nanomolar concentrations of ET-1 and TNF-α for long periods of time (above 4 h) increased Pgp-mediated transport compared to control levels, and after a 6-h exposure Pgp transport was roughly doubled (Bauer et al. 2007). Von Wedel-Parlow et al. (2009) reported that Pgp levels were increased by TNF-α within 6 h but decreased later (Von Wedel-Parlow et al. 2009). Poller et al. reported similar results working with a human cell line of immortalized brain microvessels endothelial cells; they also noted that IL-6 treatment produced a slight decrease in Pgp mRNA expression (Poller et al. 2010). Regarding the influence of Reactive Oxygen Species on efflux transporters expression levels, the first evidence of up-regulation of Pgp came from in vitro experiments on primary culture of rat brain endothelial cells (Felix and Barrand 2002). Four hours after exposure to 100 μM H2O2, up-regulation of Pgp was observed at both mRNA and protein levels, which continue to increase up to a maximum at 48 h. A biphasic up-regulation was also observed after a 6-h hypoxia and subsequent reoxigenation (H/R) treatment; in this case, return to basal levels was observed following reoxigenation by 48 h. More recently, Robertson et al. (2009) reproduced the previous experiments comparing the effects of H2O2 H/R treatments in primary rat brain endothelial cells and immortalized rat brain endothelial cells. Although the production of Reactive Oxygen Species after H2O2 was more pronounced in immortalized cells lines, similar up-regulation of Pgp, at the protein level, was observed after the oxidative stress treatments in both types of cells. Similar results were obtained with other models, such as exposure to diesel exhaust particles or glutathione depletion (Hartz et al. 2008; Hong et al. 2006; Wu et al. 2009).

The discovery of the role of pro-inflammatory mediators and oxidative stress in epilepsy explains current interest in immune, antiinflammatory and neuroprotective therapies as potential strategies to improve disease prognosis. For example, it was observed that ascorbic and lipoic acids ameliorate oxidative stress in experimental seizures (Santos et al. 2009; Militão et al. 2010). ACTH—a peptide that releases endogenous steroids in the patient—is used as a treatment for infantile spasms, a childhood refractory epilepsy; its efficacy has been confirmed in controlled trials (Pellock et al. 2010), while the use of other anti-inflammatory therapies such as steroids remains controversial due to current lack of controlled clinical studies (Vezzani et al. 2011).

14.5 Conclusions

There are currently four different hypotheses for drug resistant epilepsy. None of them seems to completely explain all cases of refractory epilepsy, but subgroups of unresponsive patients instead. At first sight, each of them claims for a different therapeutic approach. Among the strategies proposed to overcome transporter-­mediated refractory epilepsy, computer-aided research on new AEDs which are not recognized by ABC transporters, and circumventing transport by either prodrug design or nanoscale drug carriers seem as the best alternatives. Considering the efflux transporters’ role in the disposal of potentially toxic endogenous and exogenous compounds, we do not believe adjuvant inhibitory therapies as a feasible option in the case of long-term treatments (e.g., AEDs). Still, one should consider that inhibition of a given transporter is often compensated by up-regulation of another member of the ABC superfamiliy. Regarding the target hypothesis, design of multi-target agents that introduce mild perturbations to several AED targets seems to be a good alternative for the treatment of those patients with certain altered, unsensitive target. Finally, considering the intrinsic severity hypothesis, and since inflammation and oxidative stress seem to have a role in generation and exacerbation of seizures, controlled trials on the possible effects of antioxidants, immune and anti-inflammatory medication on epilepsy may have an impact on disease prognosis and severity, and consequently improve the chance of seizure remission.

Recent findings on the effect of oxidative stress and inflammation on ABC transporters expression confirm the idea that some (if not all) of the hypothesis of drug resistant epilepsy can be integrated. More research on the relationship between oxidative stress and alterations to AED targets should be explored. Revealing the fine mechanisms that govern biochemical pathways and cellular events involved in epileptogenesis (e.g., angiogenesis, inflammation) would create new opportunities for the development of innovative antiepileptic medications.