Benzodiazepine potency correlates with binding affinity at the benzodiazepine sites on neuronal GABA_A receptors (Table 58.1) (14,15). An electron-withdrawing group at the 7 position (chloride for diazepam, lorazepam, and clorazepate; a nitro group for clonazepam and nitrazepam) increases receptor binding affinity (16) and especially potency; all useful anticonvulsant benzodiazepines have such a group. Reducing the 7-nitro substituent of clonazepam to produce its less active metabolite (7-amino-clonazepam) greatly decreases receptor-binding affinity and anticonvulsant potency. A methyl group on the nitrogen at position 1 (as in diazepam and clobazam) increases binding affinity and potency; a hydroxyl group at position 3 (as in lorazepam) decreases potency and binding affinity. Several benzodiazepines have a halogen at the 2′ position (chloride for lorazepam and clonazepam) that increases receptor-binding affinity and potency. By itself, however, relative potency is not indicative of a drug’s anticonvulsant selectivity or therapeutic usefulness.

Activity at the GABA_A receptor is a function of the drug’s affinity for the benzodiazepine binding site and its intrinsic allosteric effect on that receptor. Individual compounds have widely variable efficacy. Benzodiazepines used as AEDs are thought to be full agonists that maximally enhance GABA_A receptor activity. Competitive antagonists bind to the benzodiazepine site but do not affect GABA_A receptor function. The antagonist, flumazenil, is used to reverse benzodiazepine-induced sedation in anesthesia (17,18) and to treat benzodiazepine overdose (19). Several “partial agonists” at the benzodiazepine binding site—including abecarnil (20), imidazenil (21), and bretazenil
(22)—although less effective than full agonists like diazepam, are anticonvulsant in animal models and appear less likely to promote tolerance (23,24). Other “inverse agonists” at the benzodiazepine site, including some β-carbolines, inhibit GABA binding or GABA-evoked currents (25). They can induce convulsive seizures or anxiety (25,26) but as yet have no clinical utility.

Benzodiazepines are effective against virtually every type of experimental seizure, but individual drugs show large quantitative differences for specific seizure models and other clinical effects (27). Benzodiazepines are particularly effective against seizures induced by the convulsant pentylenetetrazol, often considered a model of generalized absence seizures (28), but are less useful against tonic seizures induced by “maximal electroshock” (29), a model of generalized tonic-clonic convulsions. Relative doses necessary to achieve a particular effect also differ. The diazepam dose to block pentylenetetrazol seizures is 1% of that to abolish the righting response; for clonazepam, it is less than 0.02%, suggesting a larger therapeutic window. However, the dose of diazepam necessary to block maximal electroshock seizures is 11-fold higher than the dose required to suppress pentylenetetrazol seizures but is 2000-fold higher for clonazepam (27); this suggests that diazepam is more useful against generalized tonic-clonic seizures. Diazepam (30) and clorazepate (31) also slow the development of kindling, an animal model in which repeated subconvulsive electrical stimulation produces increasingly severe seizure activity (32).

The association, or “coupling,” between benzodiazepine binding and GABA binding (33), demonstrated in early binding studies, suggested activity at a specific receptor associated with the bicuculline-sensitive GABA_A receptor to modulate its actions on a chloride channel (33, 34, 35). Thus, a “GABA_A receptor complex” appeared to incorporate binding sites for GABA, benzodiazepines, and barbiturates with a ligand-gated chloride channel. Electrophysiologic studies demonstrated that benzodiazepines increased the amplitude of GABA-mediated inhibitory postsynaptic potentials (IPSPs) (12) by raising the opening frequency of the GABA-gated chloride channel (36). This mechanism was later confirmed with single-channel studies (37).

In whole-cell patch-clamp recordings of CNS neurons, benzodiazepines shift the concentration-response curve
for GABA leftward (Fig. 58.2), increasing current amplitudes at lower GABA concentrations but not increasing maximal current (38). This shift is caused by an enhanced affinity for GABA at its binding site, with no change in channel-gating kinetics (37). The benzodiazepines thus increase the current produced by low GABA concentrations but have no effect at high GABA concentrations, at which receptor binding is saturated. Studies of GABAergic inhibitory postsynaptic currents (IPSCs) have suggested that GABA is present in the synaptic cleft for 1 to 3 ms at high concentrations (about 1 mM) (39,40). Thus, at individual synapses, benzodiazepines prolong the mIPSC decay phase (41,42) by slowing the dissociation of GABA from the receptor (43,44) without changing maximal mIPSP amplitude. Prolongation of the mIPSC increases the likelihood of temporal and spatial summation of multiple synaptic inputs, which, in turn, raises the amplitude of stimulus-evoked IPSCs. The benzodiazepines thus increase the inhibitory “tone” of GABAergic synapses, which prevents or limits the hypersynchronous firing of neuron populations that underlies seizure activity (45).

The anticonvulsant properties of benzodiazepines are likely related to the prominent role of GABA_A receptors in epilepsy. Substantial evidence links epilepsy with dysfunction of GABAergic inhibition (45). GABA_A receptors are the target not only of the benzodiazepines but also of the barbiturates and, indirectly, of tiagabine and vigabatrin, which increase GABA concentration at the synapse (45). Several animal models of epilepsy demonstrate altered numbers or functions of GABA_A receptors (46). Moreover, changes in the composition or structure of the transmembrane protein subunits that make up GABA_A receptors can result in epilepsy. Expression of GABA_A receptor subunits is altered in the hippocampi of experimental animals with recurrent seizures (47) and in patients with temporal lobe epilepsy (48,49). Mice lacking the GABA_A receptor β₃ subunit have seizures and behavioral features of Angelman syndrome (50), a neurodevelopmental disorder associated with severe mental retardation and epilepsy caused by a mutation affecting the β₃ subunit on chromosome 15q11-13 (51). In addition, two mutations in the γ₂ subunit that impair GABA_A receptor function (52), K289M (53) and R43Q (54), have been linked to a syndrome of childhood absence epilepsy and febrile seizures; a loss-of-function mutation in the α₁ subunit was found in a family with autosomal dominant juvenile myoclonic epilepsy (55).

With a few exceptions (26), the benzodiazepines derive anticonvulsant properties from specific interaction with GABA_A receptors. In one study, their high-affinity interactions with GABA_A receptors did not fully explain the anticonvulsant effect, part of which occurred at concentrations much higher than necessary to saturate the GABA_A receptor benzodiazepine binding site, was exponential rather than saturable, and was not antagonized by flumazenil (56). Benzodiazepines, but not GABA or muscimol, were anticonvulsant when injected into the substantia nigra pars reticulata (57,58). Pharmacodynamic issues in benzodiazepine metabolism may explain these findings (see p. 834); alternatively, other sites of action may be involved. At doses used to treat status epilepticus, benzodiazepines also inhibit voltage-gated sodium (59) and calcium channels (60) and increase GABA levels in cerebrospinal fluid (61). Benzodiazepines, however, have no interaction with the G-protein-linked GABA_B receptor, which can either suppress voltage-gated Ca²⁺ channels or activate inward-rectifying K⁺ channels (62).

Benzodiazepines also bind to the “peripheral benzodiazepine receptor” (PBR) (63,64), an 18-kDa protein that functions as part of the mitochondrial permeability transition pore involved in cholesterol transport (64), apoptosis, and regulation of mitochondrial function (65). Although the PBR is widely expressed throughout the body, in the central nervous system (CNS) it is restricted to ependymal cells and glia (63) and is therefore unlikely to have a role in the clinical properties of benzodiazepines.

Benzodiazepine enhancement of GABA_A receptor function may not always be anticonvulsant or even inhibitory. Early in CNS development, neurons do not express the major chloride transporters (e.g., the potassium-chloride cotransporter, KCC2) that lower intracellular chloride concentration and create the negative driving force for chloride ions found in adult neurons. As a result, activation of GABA_A-receptor channels can be depolarizing and excitatory and may play a trophic role in neuronal morphogenesis or synaptogenesis, or both (66,67). In fact, endogenous GABA appears to be proconvulsant in early postnatal rat hippocampal slices, as GABA_A antagonists blocked epileptiform activity induced by depolarization with high external [K⁺] (68). However, benzodiazepine efficacy appears to be intact, likely because persistent opening of GABA channels (in the presence of benzodiazepines) may reduce the depolarizing chloride reversal potential, resulting in “shunt” inhibition or subthreshold depolarization that inactivates sodium channels and prevents firing of action potentials (69,70).

The current through GABA_A receptor channels can be altered by changes in concentration of intracellular bicarbonate (71,72), which also flows though the channel (73). These changes may underlie a reduction in synaptic GABA current amplitudes during development of benzodiazepine tolerance (74,75). Depolarizing GABA_A currents may also be a source of interictal spike activity, as observed in epileptic subiculum neurons in hippocampal brain slices
removed from patients with temporal lobe epilepsy (76). Changes in the GABA current reversal potential might also explain why diazepam can be less effective in children with epileptic encephalopathies (77), and why, rarely, it can cause status epilepticus in patients with Lennox-Gastaut syndrome (78,79).

The GABA_A receptors have binding sites for agents that modulate receptor function, including the benzodiazepines, barbiturates, neurosteroids, general anesthetics, the novel anticonvulsant loreclezole, and the convulsant toxins picrotoxin and bicuculline. Recent research into the molecular biology of GABA_A receptors has clarified the mechanisms of benzodiazepine action.

GABA_A receptors belong to the ligand-gated channel superfamily that includes the nicotinic acetylcholine, glycine, and the serotonin 5HT₃ receptors (80). GABA_A receptors are pentameric (81) transmembrane chloride channels assembled from combinations of protein subunits from several families (Fig. 58.2). Seven subunit families have been identified (12), with 30% to 40% homology between families and about 70% homology within families. In mammals, 16 subunit subtypes have been cloned, including 6 α, 3 β, and 3 γ subtypes, single members of the δ, π (82), and ε (83) families [θ (84) may be the fourth member of the β family (85)], as well as alternatively spliced variants of the β₂ and γ₂ subtypes. Random pentomeric combinations of all of the subunits would produce tens of thousands of subunit compositions. Because GABA_A receptor subtypes are differentially expressed by CNS region and cell type (86) and are developmentally regulated (87,88), the total number of possible isoforms in specific brain regions and individual neurons is reduced. The most common GABA receptor conformation contains the α₁, β₂, and γ₂ subtypes, with presumed stoichiometry of 2α, 2β and a single γ subunit; the δ subunit may in some cases substitute for γ. The subunits are arranged around a central water-filled pore that can open to conduct Cl⁻ ions when GABA is bound (Fig. 58.2). Studies of recombinant receptors have shown that individual subunit subtypes confer different sensitivities to GABA_A-receptor modulators including benzodiazepines (89,90), loreclezole (91), and zinc ions (92).

Benzodiazepine augmentation of GABA_A receptor currents requires a γ subunit, and the specificity of benzodiazepine responsiveness is determined by which α subunit is present (12,93). The effect of GABA_A receptor subunit composition on benzodiazepine binding has been well characterized by radioligand binding studies (94) and electrophysiology of the recombinant receptors expressed in fibroblasts (85,95). The α1 subunit results in a receptor with high affinity for the imidazopyridine hypnotic zolpidem, defining the “BZ-1” (or Ω-1) receptor type (94,96). The α₂ and α₃ subunits, combined with β and γ, result in BZ-2 receptors with moderate zolpidem affinity. GABA_A receptors with the α₅ subunit or the γ₃ subunit are sensitive to diazepam but have essentially no affinity for zolpidem and are called BZ-3 receptors. GABA_A receptors with the α₄ or α₆ subunits are insensitive to most benzodiazepines. Given the dependence of benzodiazepine binding and action on α and γ subunits, the benzodiazepine binding site is, not surprisingly, in a cleft between them (97).

Exploration of the roles of the α subunits has been based on the discovery of a single histidine (H) residue found in all benzodiazepine-sensitive α subunits (H101 in the rat α₁ subunit), but not in the benzodiazepine-insensitive α₄ or α₆ subunits. Mutation of the arginine (R100) to H in α₆ dramatically increases benzodiazepine binding in this normally insensitive subunit, while mutation of H101 to R in α₁ reduces benzodiazepine sensitivity (98); this H residue was discovered in a strain of “alcohol-nontolerant” rats that had a point mutation in the α₆ subunit (R100Q). This spontaneous mutation made their α₆-containing GABA_A receptors (found mostly in the cerebellum) sensitive to diazepam and likely accounted for their ethanol and benzodiazepine intolerance (99). Subsequent studies have used mice carrying “knock-in mutations” with individual α subunits mutated to be benzodiazepine insensitive. In homozygous α₁ (H101R) knock-in mice, the benzodiazepine anxiolytic effect was intact, but the drug did not protect against pentylenetetrazol-induced convulsions and did not produce sedation or amnesia; therefore, binding to the (wild type) α₁ subunit may be responsible for sedative, amnestic, and anticonvulsant actions (100,101). Moreover, the sedative-hypnotic zolpidem showed no sedative effect in α₁ (H101R) mice (102). Unfortunately, these findings underscore the association between sedative and anticonvulsant efficacy for the benzodiazepines at α₁-containing GABA_A receptors. Corresponding knock-in mutations of the α₂ and α₃ subtypes (α₂[H101R] and α₃[H126R])] suggested that anxiolytic (103) and myorelaxant (104) properties of benzodiazepines derive from α₂– and α₃-containing GABA_A receptors; the α₅ subunit is critical for amnestic effects (105).

Further underscoring the role of individual subunits in GABA_A receptor function were studies that used antisense oligodeoxynucleotides (ASO) to selectively reduce the expression of specific subunits. Progesterone withdrawal results in anxiety and increased seizure susceptibility associated with an increased expression of the benzodiazepine-insensitive α₄ subunit (106). Pretreating rats with an ASO against the α₄ subunit prevented the increase in seizure susceptibility (107); this finding may have significance for catamenial epilepsy. Treatment with an ASO for the γ₂ subunit increased the convulsive threshold dose for methyl-β-carboline-3-carboxylate, a benzodiazepine inverse agonist, but not for picrotoxin or strychnine (108). ASO γ₂ treatment also reduced benzodiazepine binding but not binding to the GABA recognition site (109), indicating a reduction in benzodiazepine binding sites without a change in the number of GABA_A-receptor complexes. Similarly, mice lacking the γ₂ subunit expressed GABA_A receptors with almost no benzodiazepine recognition sites and only a minor reduction in GABA binding sites (110).

Long-term benzodiazepine treatment gives rise to tolerance, decreases in sedative or anticonvulsant properties, and dependence, the continued need for drug to prevent a withdrawal syndrome (111). Tolerance requires escalation of drug doses and increases the risk of withdrawal seizures. Long-term use of benzodiazepines can also reduce their subsequent effectiveness in acute conditions (112), such as status epilepticus. Withdrawal symptoms typically involve exacerbation of the initial anxiety, insomnia, or seizures and are more common with short-acting than long-acting agents. Rebound symptoms typically return to baseline within 1 to 3 weeks after discontinuation of the drug (113). In animal studies, tolerance develops proportionally to agonist efficacy. Partial agonists produce much less tolerance than full agonists, and the antagonist flumazenil causes no tolerance-related changes in receptor number or function (24). As tolerance to one benzodiazepine may not induce tolerance to a different one, drug-specific interactions at their receptors may be operative (114). The duration of tolerance also varies among benzodiazepines (115). Several studies have noted changes in expression of GABA_A receptor subunits (116, 117, 118) as well as altered synaptic physiology (119,120) that depend not only on the drug and dosage but also on the duration and method of administration; all contribute to benzodiazepine receptor occupancy. Measurements of tolerance also vary by seizure model (pentylenetetrazole, bicuculline, pilocarpine, kindling, etc.) and by the behavioral tests used to assess the other benzodiazepine clinical properties (121).

Flumazenil (RO15-1788) binds to the benzodiazepine site without changing GABA site binding or GABA-evoked currents; it thus meets the pharmacologic definition of an antagonist. Used primarily to reverse benzodiazepine-induced sedation (17,18), flumazenil may also reverse hepatic coma (122,123) in patients naive to benzodiazepines; this bolsters arguments for an endogenous benzodiazepine ligand or “endozapine” displaced by flumazenil (124). The role of an endogenous “diazepam binding inhibitor” peptide (125) in inhibitory neurotransmission remains unclear.

Brief exposure to flumazenil can reverse tolerancerelated changes in GABA_A receptor function (126,127) and subunit expression (128). The use of intermittent low doses was explored in three patients with daily seizures who had become tolerant to clonazepam (1 mg twice daily) (129). A single intravenous dose of flumazenil (1.5 mg) resulted in mild shivering lasting 30 minutes, followed by seizure freedom for an average of 13 days. Refinement of this approach may extend the use of benzodiazepines as long-term therapy for epilepsy.

Curiously, flumazenil has shown anticonvulsant efficacy in some animal models, possibly as a result of partial agonism at high doses (130,131) or antagonism of an endogenous proconvulsant (129). Epileptiform discharges in hippocampal slices were also reduced (132), and kindling was slowed (133). Doses from 0.75 to 15 mg suppressed focal epileptiform activity in six patients with temporal lobe seizures but did not affect generalized spike-and-wave activity in six patients with generalized seizures (134). Several small studies have suggested possible benefit as an AED in humans (134,135). Orally administered flumazenil reduced seizure frequency by 50% to 75% in 9 of 11 previously untreated individuals with epilepsy and in 9 of 16 patients when used adjunctively (136). Prevention of interictal epileptiform discharges on the electroencephalogram (EEG) was similar to that of diazepam (129,137).

Flumazenil can precipitate seizures in patients with hepatic encephalopathy or benzodiazepine dependence or in patients who have attempted overdose (e.g., with tricyclic antidepressants) (138). Flumazenil has been administered to patients previously treated with benzodiazepines to precipitate partial seizures during inpatient localization of seizure onset (139). [11C]Flumazenil has been used in positron emission tomography studies to demonstrate regions of neuronal loss associated with epilepsy (140,141) and may aid in localizing the seizure focus in patients with dual pathologic conditions (142).

A cause of significant morbidity and mortality (178,179), status epilepticus requires urgent treatment to avoid neuronal damage and its neurologic consequences (180,181). The benzodiazepines have become initial therapy for status epilepticus owing to their rapid onset and proven efficacy (182), with relatively minor cardiotoxic reactions or respiratory depression compared with barbiturates (183). In the multicenter, double-blind Veterans Affairs Cooperative Status Epilepticus Trial (184), patients were randomized to receive lorazepam (0.1 mg/kg), phenytoin (18 mg/kg), phenobarbital (15 mg/kg), or diazepam (0.15 mg/kg) followed by phenytoin (18 mg/kg). Lorazepam was effective in 64.9% of patients, phenobarbital in 58.2%, diazepam/phenytoin in 55.8%, and phenytoin in 43.6% (p <0.001 lorazepam vs. phenytoin alone). Efficacy appeared to correlate with the rate at which therapeutic drug concentrations were achieved; lorazepam required the shortest time, phenytoin the longest (p <0.001).

Lorazepam and diazepam were compared for treatment of simple partial, complex partial, and secondarily generalized tonic-clonic status epilepticus in a double-blind study of 78 adults with symptomatic localization-related epilepsy (182); idiopathic generalized epilepsy with absence status epilepticus was also represented. After the first injection, lorazepam (4 mg) stopped status epilepticus in 78% of patients and diazepam (10 mg) in 58%; effectiveness was 89% and 76% after the second injection (p = NS between drugs). An open-label, prospective, randomized trial compared lorazepam (0.05 to 0.1 mg/kg) and diazepam (0.3 to 0.4 mg/kg) in children with acute convulsions including convulsive status epilepticus (185). Lorazepam was statistically more effective (p <0.01) after the first dose and apparently safer. Its superior efficacy may be the result of a longer duration of action, based on a longer distribution half-life.

Lorazepam may replace diazepam for prehospital treatment of status epilepticus. A large clinical comparison found that status epilepticus had terminated by the time of arrival at the emergency department in 59.1% of patients treated with lorazepam (2 mg), in 42.6% treated with diazepam (5 mg), and in 21.1% given placebo (186). Rates of circulatory or ventilatory complications were similar (10.6 for lorazepam, 10.3% for diazepam) and lower than those of placebo (22.5%).

Early treatment increases the probability of seizure termination (186), most likely because prolonged seizures alter GABA_A receptor susceptibility to benzodiazepines (187). Reduction in benzodiazepine sensitivity of the GABA_A receptor can occur within minutes in status epilepticus (188,189) and may be responsible in part for both the persistent epileptic state and its refractoriness. Refractoriness to diazepam may be mediated by N-methyl-D-aspartate (NMDA) receptor mechanisms, as NMDA antagonists improve the response to diazepam in late pilocarpine-induced status epilepticus (190). Moreover, NMDA receptors are upregulated during benzodiazepine tolerance, and NMDA antagonists (e.g., MK801) can block benzodiazepine-withdrawal seizures (191). These findings suggest a possible strategy for treatment of late benzodiazepine-refractory status epilepticus with combinations of a benzodiazepine (e.g., intravenous midazolam) and an NMDA-receptor antagonist such as the dissociative general anesthetic, ketamine. A recent trial of oral ketamine for refractory nonconvulsive status epilepticus showed efficacy in all five children studied (192). Such approaches will require validation in controlled clinical trials.

Both lorazepam and diazepam have been approved by the U.S. Food and Drug Administration (FDA) for treatment of status epilepticus in adults; diazepam has also been approved in children more than 30 days old. Parenteral preparations of midazolam, flunitrazepam, and clonazepam expand the treatment possibilities. Clonazepam for parenteral administration is currently marketed only in Germany and the United Kingdom, and flunitrazepam is not available in the United States. Intramuscular injection and intranasal (153,154), buccal (151), endotracheal (193,194), or rectal (149,195,196) instillation also rapidly produce therapeutic levels and are effective against status epilepticus or seizure clusters.

The need for immediate high drug levels to handle serial seizures is less urgent, and ease of administration by family or allied health workers becomes important. Diazepam rectal gel prevents subsequent seizures during seizure clusters (195,196) and can reduce the frequency of emergency department visits (145). Table 58.2 compares the clinical and pharmacologic properties of the benzodiazepines used for acute seizures. Individual agents can be selected for specific clinical situations. Repeated seizures in a patient rapidly weaned from anticonvulsants for inpatient monitoring could be treated with diazepam (rather than lorazepam), as its shorter peak duration of action may be less likely to suppress seizure activity needed later.

Although their long-term use in epilepsy is limited by sedation and tolerance, benzodiazepines may figure in the adjunctive treatment of myoclonic and other generalized seizure types, or seizures with comorbid anxiety disorders. For example, lorazepam improved control of seizures associated with psychological stressors (197). The benzodiazepines can be an integral part of a rational polypharmaceutical regimen, defined as the minimum effective AED combination for seizure control.

Intermittent administration when seizure thresholds are transiently reduced may be the ideal strategy for benzodiazepines. Not only are they suited pharmacokinetically for such an application, but such short-term use also may avoid tolerance. For example, catamenial seizures improved with intermittent use of clobazam (198). Studies
of efficacy for specific indications with the individual agents are discussed below.