PHARMA OKINETICS -THERA PEUTICS

elin. Pharmacokinet. 23 (3): 216-230, 1992 0312-5963/ 92/0009-0216/$07.50/0 © Adis International Limited. All rights reserved. CPK1211

Pharmacokinetic Optimisation of Anticonvulsant Therapy Alison H. Thomson and Martin J. Brodie University Department of Medicine and Therapeutics, Western Infirmary, Glasgow, Scotland

Contents 216 219 219 221 221 223 223 223 224 224 224 224 225 225 225 226 226 226 226 227 227 227 227 227 228

Summary

Summary I. Starting Anticonvulsant Therapy 1.1 Monotherapy 2. Maintenance Dosage Adjustment 2.1 Monotherapy 2.2 Polypharmacy 3. Status Epilepticus 3.1 Phenytoin 3.2 Phenobarbital 4. Specific Patient Groups 4.1 Neonates 4.2 Children 4.3 Hypoalbuminaemia 4.4 Renal Impairment 4.5 Hepatic Impairment 4.6 Critically III Patients 4.7 Pregnancy 4.8 The Elderly 4.9 Overdosage 5. New Antiepileptic Drugs 5.1 Vigabatrin 5.2 Lamotrigine 5.3 Oxcarbazepine 5.4 Gabapentin 6. Conclusions

Changing attitudes towards the use of antiepileptic drugs have led to an emphasis on monotherapy with serum concentration measurement coupled with standard, weight-adjusted starting and maintenance regimens to guide initial therapy and subsequent dosage titration. Currently, the established anticonvulsants are carbamazepine, valproic acid (sodium valproate) and phenytoin. Phenobarbital is now less commonly prescribed due to its propensity to produce sedation and impair cognitive function. The value of pharmacokinetic optimisation with valproic acid is limited by its wide therapeutic index, large fluctuations in the concentration-time profile and concentration-dependent protein binding. Thus, although serum concentrations are often measured, they are rarely sub-

Anticonvulsant Optimisation

217

jected to pharmacokinetic interpretation. Carbamazepine has a flatter concentration-time profile than valproic acid. Its target range is more clearly defined and it undergoes autoinduction of metabolism and interacts with other drugs. Pharmacokinetic principles can, therefore, be more readily applied to carbamazepine, although, in general, a simple clinical approach to its use is usually satisfactory. Phenytoin has required the greatest pharmacokinetic input due to its nonlinear pharmacokinetics and narrow target range. Many different graphical methods, equations and computer programs have been reported, some of which dema.nd 2 steady-state, dose-concentration pairs; others function satisfactorily with only I. Recent attempts have been made to interpret non-steadystate data. In addition, a number of workers have demonstrated the value of altering the population parameter estimates to account for ethnic differences. A pharmacokinetic approach can also be used to tailor the use of phenytoin in the treatment of status epilepticus. Dosage alterations may be needed for specific patient groups. In particular, children generally require higher dosages on a weight-for-weight basis than adults, while equivalently lower dosages should be given to neonates. Most anticonvulsants are principally cleared by hepatic mechanisms, so dosage adjustment is not usually required in renal disease, although care must be taken in interpreting serum concentrations because of changes in protein binding. Close monitoring is required in the elderly and patients with hepatic impairment, while increased dosages may be needed in critically ill patients and during pregnancy. Pharmacokinetic principles can be used in the treatment of treat self-poisoning with anticonvulsants. There are few data available on the pharmacokinetics of vigabatrin, lamotrigine, oxcarbazepine and gabapentin in patients. Due to its mode of action in binding irreversibly to its target enzyme, serum concentration monitoring of vigabatrin plays no role in optimising therapy. The value of applying pharmacokinetic principles with the other 3 drugs remains to be investigated. Of these, lamotrigine seems the most likely candidate for a pharmacokinetic approach.

The management of epilepsy has been greatly influenced by the introduction of rapid and reproducible antiepileptic drug assays. This had led to more rational use of this group of therapeutic agents, with a focus on optimising therapy by integrating drug concentration measurements with clinical observations. This article outlines the pharmacokinetic approaches which can be used to guide dosage adjustment with established antiepileptic drugs and discusses the evidence for a doseresponse-toxicity relationship with some of the newer anticonvulsants. Carbamazepine, valproic acid (sodium valproate) and phenytoin are the major drugs used to treat epilepsy. Phenytoin can now perhaps be regarded as second-line therapy because of its potential to cause cosmetic and psychosocial side effects and the problems of dosage optimisation without concentration monitoring (Brodie 1990a). Ethosuximide is still widely used to treat absence seizures. Phenobarbital is less commonly prescribed nowadays due to its propensity to produce sedation and impair cognitive function. Nevertheless,

it has a continuing role in the management of partial and generalised seizures, particularly in neonates, and is still widely prescribed in developing countries because it is easy to use and cheap (Brodie 1990a). There is little clinical place now for primidone, which largely depends on metabolism to phenobarbital for its clinical action but which is more toxic than its active metabolite (Mattson et al. 1985). The benzodiazepines, c10nazepam and c1obazam, are mainly employed as adjunctive therapy. The former is particularly sedative and tolerance may develop to either (Brodie 1990a). Their use is largely empirical and they will not be discussed further in this review. Vigabatrin has been introduced across Europe, and lamotrigine has been licensed recently in the United Kingdom and Ireland as adjunctive (add-on) therapy when conventional treatment has failed to achieve adequate seizure control. Vigabatrin is particularly effective for partial seizures (Grant & Heel 1991), while lamotrigine, although also useful for partial seizures, is promising in primary generalised epilepsies (Brodie 1992a).

C/in. Pharmacokinet. 23 (3) 1992

218

Table I. Pharmacokinetics of commonly used anticonvulsant drugs

Drug

Absorption (F)[%]

Vd (L/kg)

PB(%)

tv. (h)

Route(s) of elimination

Comment

0.8-1.6

70-80

Clobazam

Rapid (90-100) 0.7-1.6

87-90

24-458 8-24b 10-30

Enzyme Induced autoinduction of metabolism Tolerance can occur

Clonazepam

Rapid (80-90)

2-6

80-90

30-40

Hepatic metabolism Active metabolite Hepatic metabolism Active metabolite Hepatic metabolism

Ethosuximide

Rapid (90-95)

0.6-0.9

0

20-6OC 40-6()d

Lamotrigine

Rapid (100)

1.1-1.3

55

18-30

Phenobarbital

Slow (90-100)

0.6-1.0

48-54

72-144

Phenytoin

Slow (85-95)

0.5-0.7

90-93

9-40

Primidone

Rapid (90-100) 0.4-1.0

20-30

3-19

Valproic acid (sodium valproate) Vigabatrin

Rapid 8 (100)

0.09-0.17

88-92

7-17

Rapid (60-SO)

0.6-1.0

0

5-7

Carbamazepine Slow (75-85)

Hepatic metabolism 10-20% excreted unchanged Hepatic metabolism Mainly glucuronidated

Sedative Tolerance can occur Faster CL in children

Metabolism induced or inhibited by other anticonvulsants Hepatic metabolism Enzyme induced 25% excreted unchanged Tolerance can occur Saturable hepatic Enzyme inducer metabolism Concentration-dependent tv. Hepatic metabolism Phenobarbital a major Active metabolites metabolite 40% excreted unchanged Tolerance can occur Hepatic metabolism Enzyme inhibitor Active metabolite Concentration-dependent PB Largely excreted unchanged

Long acting due to irreversible binding to GABA transaminase

a Single dose. b Multiple doses. c Children. d Adults. e Enteric-coated tablets may have a lag of 3-6h followed by rapid absorption. Abbreviations: F = bioavailability; Vd = volume of distribution; PB = protein binding; tv. = elimination half-life; CL = clearance.

The pharmacokinetics of most of these drugs are similar (table I). They are well absorbed orally, bind to circulating serum albumin (except ethosuximide and vigabatrin) and largely eliminated by hepatic metabolism (except vigabatrin). Carbamazepine, phenytoin, phenobarbital and primidone induce hepatic monooxygenase and, to a lesser extent, conjugating enzymes, while valproic acid acts as an enzyme inhibitor (Mcinnes & Brodie 1988). This leads to a number of drug-drug interactions, the results of which vary widely among patients and consequently can be difficult to predict (Brodie 1992b; Nation et al. 1990a,b; Pisani et al. 1990). Combining antiepileptic drugs invariably results,

therefore, in a panoply of pharmacokinetic interactions which are, in themselves, a good reason for preferring monotherapy whenever possible. Close relationships between measured serum concentrations and therapeutic and toxic effects are well recognised with some antiepileptic drugs (e.g. phenytoin, carbamazepine), poor with others (e.g. valproic acid, ethosuximide, phenobarbital) and nonexistent with vigabatrin. Clinical evidence supporting proposed target ranges (table I) for the established anticonvulsants was recently discussed by Choonara and Rane (1990). Such ranges may be useful as a guide to initial therapy and dosage adjustment, particularly in patients whose seizures are

219

Anticonvulsant Optimisation

infrequent. However, the anticonvulsant response, symptoms and signs of toxicity, administration history and sampling details must all be taken into account when interpreting serum concentration measurements (Brodie & Feely 1988); dosage should never be adjusted on the basis of an observed anticonvulsant concentration alone (Larkin eta!. 199Ia).

ever possible, and maintained on the lowest effective dosage (Brodie I 990a). First-line therapy for an individual patient depends on the seizure type and the toxicity profile of suitable drugs (Chadwick I 990a). In practice, carbamazepine and/or valproic acid can be regarded as drugs of first choice for most forms of generalised, partial and unclassified seizures. Indications and starting dosages for adults and children are outlined in tables II and III.

1. Starting Anticonvulsant Therapy 1.1 Monotherapy It is now well accepted that patients should be managed with a single anticonvulsant drug, when-

1.1.1 Carbamazepine Carbamazepine should be introduced at low dosages (200mg daily in adolescents and adults) due to its potential to produce psychomotor impair-

Table II. Dosage guidelines and target plasma drug concentrations for antiepileptic drugs in adults

Drug

Indications

Starting dose (mg)

Maintenance dosage (mg)

Dosage schedule

Target range (mg/L) [,umol/L]

Carbamazepine

Partial and generalised tonic-clonic seizures Adjunctive therapy for refractory partial seizures Myoclonic and generalised tonic-clonic seizures Status epilepticus Absence seizures

100-200

400-2000

bid-qid 8

4-12 [17-50]

10

10-40

od or bid

None

0.5-1

2-8

od or bid

None

500

500-2000

od or bid

25 c 50-100

50-200c 100-400

bid

40-100b [283-708] 2-4d [7.8-15.6]

30-60

60-240

od or bid

10-40b [40-172]

200-300

100-700

od or bid

10-20 [40-80]

125-250

250-1500

bid

500

500-3000

od or bid

6-12b [25.55] 50-100b [347-693]

500-1000

2000-4000

od or bid

Clobazam Clonazepam

Ethosuximide Lamotrigine

Adjuvant therapy for partial and generalised tonic-clonic seizures Phenobarbital Partial and generalised tonic"clonic seizures Status epilepticus Phenytoin Partial and generalised tonic-clonic seizures Status epilepticus Primidone Partial and generalised tonic-clonic seizures Valproic acid Primary generalised (sodium valproate) epilepsies Partial and secondary generalised seizures Vigabatrin Adjunctive therapy for refractory partial seizures a b c d

Once or twice daily if a controlled release formulation is used. Target range unhelpful. If administered with concomitant valproic acid. Target range not validated. Abbreviations: qid = 4 times daily; bid = twice daily; tid = 3 times daily; od = once daily.

None

Clin. Pharmacokinet. 23 (3) 1992

220

Table III. Dosage guidelines for antiepileptic drugs in children

Drug

Indications

Initial dose

(mg/kg/day)

Maintenance dosage

Dosage schedule

(mg/kg/day) Carbamazepine Clobazam Clonazepam

Ethosuximide Lamotrigine Phenobarbital

Phenytoin Valproic acid (sodium valproate) Vigabatrin

a

Partial and generalised tonic-clonic seizures Adjunctive therapy for refractory partial seizures Myoclonic epilepsy Lennox-Gastaut syndrome Infantile spasms Status epilepticus Generalised absences Partial and generalised tonic-clonic seizures Lennox-Gastaut syndrome Generalised tonic-clonic seizures Newborn seizures Status epilepticus Partial and generalised tonic-clonic seizures Status epilepticus Generalised epilepsies Partial seizures West syndrome

When administered with valproic acid. = twice daily; tid = 3 times daily; od

Abbreviations: bid

5 0.25

10-25 0.25-0.5

bid or tid bid or tid

0.025

0.025-0.1

bid or tid

10 0.58 2 4

15-40 1-58 5-15 4-10

od od or bid

5

5-15

od

10

15-40

od or bid

25

25-80

od or bid

od or bid

= once daily.

ment (headache, nausea, dizziness, diplopia, ataxia) and because clearance (CL) will initially be low in the naive patient (Rapeport et al. 1983). Drug concentrations accumulate during the first week of treatment and then gradually decline over the next month (Bertilsson et aI. 1986; Macphee & Brodie 1985). This decrease is a consequence of autoinduction of metabolism, which can lead to a 3-fold increase in CL over the first 3 to 5 weeks of therapy in some patients (Eichelbaum et al. 1985). The dosage of carbamazepine may, therefore, need to be increased shortly after starting treatment, to take these changes into account (Macphee & Brodie 1985). As the dose: concentration ratio increases with age (Levy & Kerr 1988), children require higher dosages of carbamazepine on a mgjkg basis than adults (Rylance et al. 1979). Elderly patients should be started on a lower dosage (lOOmg daily) and close clinical monitoring is advised during the first

few weeks of treatment to ensure that side effects do not occur. The lower starting dosage reflects the extent of autoinduction of carbamazepine metabolism, which is likely to be less in older than in young patients (Smith et al. 1991). This policy of initial caution is prudent, as tolerance to the central nervous system (CNS) side effects of carbamazepine takes place quickly (Larkin et al. 1992). 1.1.2 Valproic Acid Yalproic acid can be introduced in adolescents and adults at the relatively higher dosage of 500mg daily (table II), which can be adjusted upwards or downwards according to clinical response or the development of side effects. Higher starting doses can cause gastrointestinal upset (Pugh & Garnett 1991). Since the drug can take several weeks to become fully effective (Rowan et al. 1979), frequent dosage adjustments shortly after starting therapy are unjustified.

Anticonvulsant Optimisation

1.1.3 Phenytoin Since phenytoin can take up to 4 weeks to reach steady-state, therapy should be started and continued with an initial maintenance dosage of around 5 mg/kg/day (tables II and III). If urgent control is required, an intravenous or oral loading dose of 15 mg/kg can be given, to be followed up with 5 mg/kg/day. The followup regimen should be carefully monitored as it will produce toxicity in a small minority of patients. 1.1.4 Ethosuximide No loading dose is required for ethosuximide, but it should also be introduced slowly (tables II and III) as the patient may experience initial gastrointestinal or eNS side effects. The dosage can be gradually increased to 20 mg/kg/day after about 7 days and then by further increments to 30 mg/ kg/day as necessary. 1.1.5 Phenobarbital To minimise sedation and allow tolerance to develop to this common side effect, phenobarbital should be started at a low dosage (30 to 60mg daily). This may be increased gradually according to clinical requirements.

2. Maintenance Dosage Adjustment 2.1 Monotherapy 2.1.1 Carbamazepine Due to auto induction of metabolism, the maintenance dosage of carbamazepine may need to be increased during the first few months of treatment if concentrations are to be maintained within the range achieved during early therapy (Macphee & Brodie 1985). The extent of auto induction differs greatly among patients (Macphee et al. 1987), making dosage requirements highly variable. Levy and Kerr (1988) have produced guidelines, based on weight and age, for the dosage which will produce a steadystate carbamazepine concentration of 6 mg/L. These are useful for initial maintenance dosage selection. Further modifications can be made as clinically indicated. Once steady-state has been achieved, the carb-

221

amazepine dosage can usually be adjusted by simple proportion according to a measured concentration, although there is some evidence that further autoinduction of metabolism can occur at higher dosages and concentrations (Battino et al. 1980; Rapeport et al. 1983). Other workers, however, have not been able to confirm this finding (Bertilsson et al. 1986) and it may be a patient-specific effect. The major metabolite of carbamazepine, carbamazepine-IO,II-epoxide (carbamazepine-E), is present at concentrations of 10 to 40% of the parent drug (Johnson & Brodie 1990). Higher ratios are associated with coadministration of other anticonvulsants (Brodie et al. 1983). Although carbamazepine-E contributes to the pharmacological and neurotoxic effects of carbamazepine (Gillham et al. 1988; Pugh and Garnett 1991; Tomson et al. 1990), this metabolite is not routinely measured. If carbamazepine is administered at 6 to 8h intervals, the sampling time is not critical because the concentration-time curve will be relatively flat. However, if the drug is only given once or twice daily, there may be large fluctuations in the concentration-time profile (Macphee et al. 1987), which can confound interpretation of serum concentrations (Larkin et al. 1991 a). This is a particular problem in patients who require high dosages of carbamazepine. Under such circumstances it is important that samples are taken at the same time each day if they are to be used as a guide to dosage (Larkin et al. 1991 a). Excessive fluctuations in the concentration-time profile can be dampened by more frequent carbamazepine administration or by using a controlled-release formulation (Bialer 1992; Larkin et al. 1989). If the latter is preferred, a slightly higher dosage may be required because bioavailability may be lower than that with conventional carbamazepine (May & Rambeck 1989; McKee et al. 1991). Pharmacokinetic interpretation of serum carbamazepine concentrations has been investigated using a Bayesian parameter estimation program (Garcia et al. 1988). The authors found this to be superior to linear regression analysis when only 2 concentration measurements were available. Both methods produced clinically acceptable precision

222

in terms of prediction of future concentrations when 3 or 4 concentrations were employed. However, in our experience, care must be taken with such programs. Carbamazepine is slowly absorbed, particularly if a controlled-release formulation is used or if dosage intervals are short, so that the assumption that absorption is complete between doses (Kelman et al. 1982) may be violated, resulting in inappropriate CL estimates. The possibility of nonlinearity in the dose-concentration relationship should also be borne in mind. 2.1.2 Vaiproic Acid Dosage adjustment of valproic acid should be made according to clinical observation rather than serum concentration monitoring, because the concentration-response-toxicity relationship is unclear (Zaccara et al. 1988) and the drug has a wide safety margin (Editorial 1988). Although 100 mgfL has been suggested as the upper limit of the target range (Vajda et al. 1978), some authors now recommend 200 mg/L as a more appropriate ceiling (Bourgeois 1989). Serum concentration measurements may be useful as a guide to compliance and in patients also receiving an enzyme-inducing anticonvulsant. There are large fluctuations in the concentrationtime profile and a significant absorption lag time of up to 6h with enteric-coated formulations (Pugh & Garnett 1991), making valproic acid concentrations difficult to interpret unless they are taken at the same time in relation to the last dose, and ideally in regimens with regular dosage intervals trough concentrations will be measured (Henriksen & Johannessen 1982). In addition, the interpretation of a measured concentration may be confounded by nonlinear protein binding which occurs at concentrations of about 100 mg/L and over (Pugh & Garnett 1991). Analysis of unbound valproic acid is not routine and a target range has not been established (Zaccara et al. 1988), although 1 to 4 mgfL has been proposed for children (Farrell et al. 1986). The recent introduction of sustained release formulations of valproic acid may reduce these problems (Bialer 1992).

Clin. Pharmacokinet. 23 (3) 1992

2.1.3 Phenytoin If seizures are well controlled and the patient is not experiencing side effects, the phenytoin concentration should be measured after 4 weeks and the dosage adjusted appropriately with followup several months later. If the patient is poorly controlled, more frequent analyses will be necessary and the maintenance dosage can be adjusted in accordance with the clinical picture. If the phenytoin concentration is below 8 mgfL, an increment of 100mg can be made; patients with concentrations between 8 and 12 mgfL can be given an extra 50mg; above 12 mgfL, the dosage should not be increased by more than 25mg increments on each occasion (Brodie 1990a). If the concentration starts to decrease after loading, a higher maintenance dosage may be required. However, many patients will be well controlled with concentrations between 5 to 10 mgfL (Woo et al. 1988). If the drug continues to accumulate, the maintenance dosage will need to be cut back. If the concentration stays reasonably constant, steady-state has probably been reached. It should be remembered that the elimination half-life (t'12) of phenytoin will lengthen as its metabolism changes from first to zero order. Optimisation of phenytoin dosage is complicated by nonlinearity in the dose-concentration relationship, which is a consequence of capacity-limited metabolism. Numerous methods have been developed to estimate the parameters Vmax (the maximum rate of metabolism) and Km (the Michaelis-Menten constant), which define the relationship between dosage and steady-state concentration. The methodology used has been reviewed recently by Pryka and colleagues (1991) and will be summarised here. Early attempts to optimise phenytoin therapy were based on linearisation of the Michaelis-Menten equation (Ludden et al. 1977; Mullen 1978), but had the disadvantage that at least 2 steady-state concentrations at 2 different dosages were required. Several methods have been proposed requiring only I steady-state concentration (Graves et al. 1986; Rambeck et al. 1979; Richens & Dunlop 1975; Wagner 1985), but the most extensively evaluated technique has been the Bayesian ap-

223

Anticonvulsant Optimisation

proach first proposed by Vozeh et al. (1981). This method has been implemented in both computerised and graphical forms and compares favourably with other methods (Armijo & Cavada 1991; Bryson et al. 1988; Chrystyn & Morgan 1986; Flint et al. 1985; Graves et al. 1986; Hudson et al. 1990; Welty et al. 1986; Yuen et al. 1983; Yukawa et al. 1988, 1989a; Zaccara et al. 1989), particularly if the population estimates are modified to account for racial characteristics of the patients under investigation (Miller & Rheeders 1989; Yukawa et al. 1989b; Yuen et al. 1989). To avoid the limitations on steady-state imposed by all these approaches, Ludden et al.( 1986) developed a program which could handle non-steady-state data. However, large errors in prediction occurred unless 3 to 5 concentrations were available. The program also had to be modified for patients receiving enteral nutrition (Godley et al. 1987; Killilea et al. 1989). Nevertheless, in a clinical study following further modification (Crowley et al. 1987), Privitera et al. (1989) found the program to be superior to physicians in achieving target concentrations of phenytoin.

2.1.4 Ethosuximide The dosage of ethosuximide can usually be adjusted empirically on the basis of normal dosage requirements (on a mg/kg basis) and clinical response. Methods for adjusting dosages to achieve target concentrations have not been established and the apparent nonlinearity of the pharmacokinetics of this drug make a pharmacokinetic approach to dosage adjustment difficult (Eadie 1976). 2.1.5 Phenobarbital If phenobarbital dosage adjustment is indicated on clinical grounds, the change can usually be calculated by proportion according to the desired target concentration. However, as there is some evidence of nonlinearity at higher concentrations (Eadie et al. 1977), it would be prudent to make more conservative dosage adjustments under these circumstances.

2.2 Polypharmacy Although around 80% of patients can be adequately controlled on anticonvulsant monotherapy (Beghi et al. 1986), some require the addition of a second drug to obtain optimal suppression of their seizures. Drug interactions can alter the dose: concentration relationship of the first drug (Brodie 1992b) and there is a significant increase in the potential for adverse, particularly sedative, side effects (Brodie et al. 1987; Gillham et al. 1988). Sedating drugs such as phenobarbital and primidone should, therefore, be avoided in such patients (Brodie 1990a). As it is difficult to predict the extent (and possibly also the direction) of these pharmacokinetic interactions, concentrations of both drugs should be monitored when a second drug is introduced and also when a second (or third) drug is withdrawn (Duncan et al. 1991). There are no clear guidelines on the ideal approach to introducing a second antiepileptic drug, but it would seem sensible to start with a similar dosage to that used in monotherapy and proceed perhaps a little more quickly with dosage increments, particularly in patients receiving enzyme-inducing anticonvulsants.

3. Status Epilepticus Patients in status epilepticus require urgent treatment. Most will respond to intravenous or rectal diazepam followed by intravenous phenytoin to prevent recurrence (Brodie 1990c). 3.1 Phenytoin If a benzodiazepine proves unsuccessful in stopping seizures, an intravenous loading regimen of phenytoin (18 mg/kg) can be tried next (Brodie I 990c). There are alternative methods, however, of administering the drug. The results of a population pharmacokinetic analysis in 49 patients by Vozeh et al. (1988) suggested that a total loading dose of 15 mg/kg divided into 3 increments of 5 mg/kg each and separated by 2h will achieve concentrations within the target range of 10 to 20 mg/L within

Clin. Pharmacokinet. 23 (3) 1992

224

6h. This regimen was tested prospectively in 40 patients (Blaser et al. 1989) and the authors found that 90% of patients had concentrations within the desired range as predicted by the model. If the patient is already taking phenytoin and a serum concentration measurement is available then, assuming a volume of distribution (V d) of 0.65 L/ kg, a suitable 'top up' dose can be calculated, as follows (Winter 1988b): D (mg) = 0.65 X weight (kg) X (C target - C measured) where C is the concentration in mg/L and D is the dose of phenytoin acid (Divide by 0.92 to correct for phenytoin sodium). 3.2 Phenobarbital Phenobarbital has lower lipid solubility and thus a slower onset of action than phenytoin, but it is an effective treatment for status epilepticus and has the advantage that it can be given intramuscularly if venous access is poor. Its main disadvantage is sedation, particularly in patients who have not previously received the drug and so have not had the opportunity to become tolerant to this side effect. A loading dose of 15 mg/kg, infused at a rate not greater than 100 mg/min, should achieve concentrations of 15 to 21 mgjL (Browne 1990; Gabor 1990). A further dose of 10 mg/kg can be given if necessary (Gabor 1990). Respiratory depression may be associated with concentrations above 35 mgfL, so intubation is a sensible precaution if additional doses are required.

4. Specific Patient Groups 4.1 Neonates 4.1.1 Phenobarbital The treatment of seizures in the neonate has focused on phenobarbital. In many cases a single loading dose of 20 mg/kg will be sufficient to achieve complete control. Maintenance dosages of 5 to 10 mg/kg/day can be continued if they are clinically indicated; the target concentration in these circumstances should be 15 to 40 mg/L. Gillman

and colleagues (1989) have stated that there is no advantage in increasing the dose above 40 mg/L and a second anticonvulsant should be added if the baby continues to have fits. In contrast, however, Crawford et al. (1988) found that refractory status epileptic us could be controlled with very high dosages of phenobarbital, which achieved concentrations from 70 to 344 mg/L. They claimed that there was no upper limit to dosage. The tl12 of phenobarbital is long (~100h), which may make multiple loading doses necessary if control is not achieved with the standard regimen, because a change in the maintenance dosage will take several weeks to exert its full therapeutic effect. Assuming a Vd of 0.9 L/kg, loading doses can be determined from a measured concentration as follows (Johannessen 1981 ): LD (mg) = 0.9 L/kg X weight (kg) X (C target - C measured) where C is the concentration in mg/L and LD is the loading dose of phenobarbital (divide by 0.91 to correct for phenobarbital sodium). The maintenance dosage can be individualised using an estimate of the elimination rate from 2 measured concentrations, as proposed by Gillman and colleagues (1983). 4.1.2 Phenytoin Phenytoin is less commonly used in the treatment of neonatal seizures than phenobarbital. The normal loading dose in neonates, 15 to 20 mg/kg, is similar to that for adults and maintenance dosage requirements usually range from 3 to 5 mg/kg/ day. Due to the lower protein binding in neonates compared with adults, the target serum concentration range should be adjusted downwards to 6 to 15 mg/L.

4.2 Children Corrected for weight, antiepileptic drugs have shorter t'l2 values in children than in adults due to a combination of lower Vd and more efficient CL. It is, therefore, important that relatively higher dosages of carbamazepine, valproic acid and

225

Anticonvulsant Optimisation

phenytoin are given, perhaps at shorter intervals, to avoid wide fluctuations in the concentration-time profile. These fluctuations can lead to toxicity at peak concentrations, breakthrough seizures when the concentration is low and difficulties in interpreting measured drug concentrations. Ethosuximide has a long t1;' of around 30h (Winter 1988a) and only needs to be given once daily. Recommended loading and maintenance doses of anticonvulsants in children are presented in table III. 4.3 Hypoalbuminaemia Since most anticonvulsant drugs are acidic or neutral, they are highly bound to serum albumin. Changes in albumin concentrations, therefore, can influence drug disposition and the interpretation of total drug concentrations. Theoretically, dosage adjustment of carbamazepine, valproic acid and phenytoin should be based on free concentrations, but in practice this is rarely done. Under these circumstances, a 'corrected' total concentration for phenytoin can be made using the following formula: C corrected = C observed + [(0.2 x albumin concentration) + 0.1] where C is the concentration in mg/L and the albumin concentration is in g/dl (Winter & Tozer 1986). 4.4 Renal Impairment Although most anticonvulsants are almost exclusively cleared by hepatic metabolism, renal failure can have an important influence on dosage optimisation. If renal function is substantially impaired, an alteration in the free fraction of phenytoin can result from a reduction in serum albumin concentrations and a decrease in the affinity of albumin for the drug. This is a particular problem in patients with creatinine clearance values below 10 ml/min (0.6 L/h) [Winter & Tozer 1986], where the free fraction can be increased by a factor of 2 to 3. These changes in binding do not affect loading or maintenance dosage requirements for

phenytoin, because the higher free fraction allows more extensive distribution and CL and so free concentrations are unchanged. However, total phenytoin concentrations will be lower and this, too, is a situation where the measurement of free drug is warranted. In patients on haemodialysis, an estimate of the 'corrected' total concentration can be made from the following equation: C equivalent = C observed + [(0.1 x albumin concentration) + 0.1] where C is the concentration in mg/L and the albumin concentration is in g/dl (Winter & Tozer 1986). Similar changes in the pharmacokinetics of carbamazepine and valproic acid are likely, but the necessity for free concentration monitoring is less apparent, particularly for valproic acid which is best monitored on clinical criteria. Vigabatrin is eliminated largely unchanged by the kidney and so the maximum dosage should be reduced in the elderly and in patients with impaired renal function (Haegele et al. 1988). In these situations it should not be necessary to exceed 2g of the drug per day. CL by exogenous removal techniques such as haemodialysis and haemofiltration are theoretically minimal because many anticonvulsants are highly protein bound and only the free drug crosses dialysis or filtration membranes. However, due to the elevated free fraction in renal failure, exogenous CL of phenytoin may be significantly increased, particularly if continuous haemofiltration is used in patients with low serum albumin levels. Dosage adjustment in this situation is best performed on the basis of free phenytoin concentrations. 4.5 Hepatic Impairment Unlike renal impairment where creatinine concentration can be used to assess dosage requirements, there is no single endogenous marker to guide dosage adjustment in patients with severe liver disease (Howden et al. 1989). All anticonvulsants, with the exception of vigabatrin, should be introduced at lower dosages than normally rec-

Clin. Pharmacokinet. 23 (3) 1992

226

ommended and increased at longer intervals, according to clinical requirements, with close clinical monitoring of the patient for the development of concentration-dependent toxicity. Hepatic impairment is often accompanied by hypoalbuminaema and so it may be useful to measure free drug concentrations in this patient group also, especially in those being treated with phenytoin. 4.6 Critically III Patients Phenytoin is the anticonvulsant of choice for critically ill patients because it can be given intravenously. Such patients may have high and changing dosage requirements (Boucher et al. 1987, 1988). In many cases the simplest approach is to give a single daily dose intravenously, based on serum concentration measurements, until seizure control is achieved. Since the serum albumin may be reduced (Boucher et al. 1987, 1988), free concentration monitoring may be useful in this clinical setting. 4.7 Pregnancy Pregnancy is associated with an increase in the CL and Vd of most anticonvulsants, which is most pronounced with phenytoin (Brodie 1990b). However, the changes in pharmacokinetics are highly variable and difficult to predict for the individual patient. Serum concentration monitoring should be performed at regular intervals, e.g. monthly throughout pregnancy and every 2 weeks in the puerperium, until dosage requirements have stabilised. More frequent analysis may be necessary if seizure control is poor or deteriorating. Adjustment can be performed using the techniques previously outlined, but it may be useful to base them on free concentrations, because pregnancy-associated hypoalbuminaemia can alter the free fraction of highly protein-bound drugs. Four phenytoin loading regimens for seizure prophylaxis in severe preeclampsia were evaluated by Ryan and coworkers (1989), who concluded that a single dose of 10 mg/kg followed by an additional 5 mg/kg 2h later produced acceptable concentra-

tions over the next 12h with minimal adverse effects in 94% of patients studied. They also suggested that patients be maintained on phenytoin for 3 to 5 days with a dosage of 200mg 3 times daily, starting 12h after loading, with adjustment as required according to phenytoin and albumin concentration measurements. 4.8 The Elderly Elderly patients are at greater risk from skeletal damage as a consequence of poor seizure control and are particularly susceptible to concentrationrelated side effects (see review by Swanson 1992). Therefore, antiepileptic drug dosages need careful titration and monitoring in this patient population. As there is enhanced pharmacokinetic variability, older patients should be started on low anticonvulsant dosages and increases should be gradual and widely spaced. Particular attention needs to be paid to the problem of compliance, because inappropriate increases in dosage may result if low concentrations are misinterpreted as indicating high CL. Proper interpretation of serum concentrations is particularly important if the patient becomes unwell, as albumin concentrations can decrease very quickly in old people who are ill. 4.9 Overdosage Pharmacokinetic principles can be employed in the treatment of self-poisoning with anticonvulsants. Vree and coworkers (1986) reported a wide variability in the carbamazepine concentration-time profile following overdosage in 5 patients. It is important to remember that further absorption can occur over a prolonged period (Durelli et al. 1989). Similar problems have been observed with phenytoin (Chaikin & Adir 1987). Treatment with multiple doses of activated charcoal can be useful in this situation, because it reduces further drug absorption and enhances elimination (Neuvonen & Olkkola 1988; Weidle et al. 1991). Drug concentrations should be monitored regularly in the early stages after overdosage, because the clinical man-

227

Anticonvulsant Optimisation

agement will be influenced by whether the concentration is rising or falling.

5. New Antiepileptic Drugs 5.1 Vigabatrin Vigabatrin is the newest antiepileptic drug to be licensed widely in Europe. It is a specific, irreversible, suicide inhibitor of "f-aminobutyric acid (GABA) transaminase, the enzyme responsible for the breakdown of the inhibitory neurotransmitter GABA (Editorial 1989). The molecule is water soluble and so the drug is excreted largely unchanged by the kidney, its CL mirroring that of creatinine (Haegele et al. 1988). Although its t'l2 is short (5 to 7h), the duration of action of a dose of the drug is >24h because it irreversibly binds to the target enzyme in the brain (Grant & Heel 1991). Accordingly, monitoring circulating concentrations of the drug is not helpful in predicting therapeutic response or avoiding dosage-related toxicity. Dosage advice is contained in tables II and III. 5.2 Lamotrigine Lamotrigine has a complicated triazine structure, which is chemically unrelated to any antiepileptic drug in current use. It has been marketed as add-on therapy (dosages in tables II and III) in refractory epilepsy in the United Kingdom and Ireland. It is thought to act by inhibiting the release of excitatory amino acids, particularly glutamate (Brodie & Porter 1990). Placebo-controlled trials of the drug as adjuvant therapy in refractory epilepsy have demonstrated undoubted efficacy (Binnie et al. 1989; Jawad et al. 1989; Loiseau et al. 1990; Sander et al. 1990) and monotherapy studies in newly diagnosed patients are under way. Following a single oral dose, lamotrigine has a t,;, of about 24h. However, treatment with other antiepileptic drugs has a substantial effect on the CL of the drug. Valproic acid inhibits its metabolism, extending the t,;, to around 60h (Binnie et al. 1986). Conversely, enzyme-inducing anticonvulsants, such as carbamazepine, phenytoin, phenobarbital and primidone accelerate its CL, yielding

t,;, values approximately half those found in healthy volunteers (Jawad et al. 1987). Accordingly, monitoring serum lamotrigine concentrations is likely to be helpful in tailoring the dosage, particularly in those patients receiving the drug in combination with 1 or more established anticonvulsants. A tentative target range of 2 to 4 mgfL has been mooted (Yuen, personal communication), although this may prove to be a little low. Clinical studies hold out promise that monitoring lamotrigine concentrations will be of practical clinical value. 5.3 Oxcarbazepine Oxcarbazepine, the 10-keto analogue of carbamazepine, has recently been marketed in Denmark. Despite their structural similarities, oxcarbazepine and carbamazepine have different metabolic profiles (Editorial 1989b). Oxcarbazepine is rapidly and extensively converted" to an active lO-hydroxy metabolite, most of which is eliminated in the urine as the glucuronide conjugate. As a consequence, oxcarbazepine may lack the general enzyme-inducing properties of carbamazepine (Larkin et al. 1991a). Another potential advantage of the new drug is the possibility of a better side effect profile; in particular, less psychomotor impairment and fewer idiosyncratic drug reactions (Dam et al. 1989). The active moiety, lO-hydroxy carbazepine, has a t,;, of around 12h, which is un. altered by multiple dose administration of the drug. Accordingly, this metabolite is a prime target for attempting to monitor the pharmacological effects and CNS toxicity of oxcarbazepine. Although little evidence is available currently to support a concentration-effect-toxicity relationship in patients with epilepsy, exploratory studies are underway in the context of controlled clinical trials. 5.4 Gabapentin Gabapentin is a GABA-related amino acid whose mechanism of action is unknown. Its t,;, is short (5 to 6h) and the drug is cleared largely unchanged by the kidney (Bartosyk et al. 1986). In an early study, a dosage-related improvement in

228

seizure control was accompanied by a linear relationship between gabapentin dosage and serum concentrations (Crawford et al. 1987). Efficacy against partial seizures has recently been demonstrated in a large, parallel group, add-on study (Chadwick 1990b). Mean and median gabapentin concentrations were higher in responders than in nonresponders. Closer analysis revealed greater efficacy with increasing gabapentin concentrations. This drug also has potential for concentration monitoring.

6. Conclusions Pharmacokinetic optimisation of anticonvulsant drugs is an attractive concept. Disappointingly, with the exception of phenytoin, which undergoes saturable hepatic metabolism, it is of limited practical use in the everyday management of patients with epilepsy. This is largely due to the interindividual variation in response and toxicity to similar concentrations of antiepileptic drugs, i.e. substantial pharmacodynamic variability. Accordingly, although it is possible with most drugs to ensure a targeted drug concentration, there is no certainty that such a concentration will provide complete seizure control without side effects. Thus, dosage adjustment in patients receiving carbamazepine, valproic acid, phenobarbital, ethosuximide and primidone is usually undertaken on the basis of clinical signs and symptoms. Of the newer drugs, lamotrigine, gabapentin and oxcarbazepine are likely to show predictable concentration-response-toxicity relationships and may prove suitable for a pharmacokinetic approach in optimising therapy.

References Armijo JA, Cavada E. Graphic estimation of phenytoin dose in adults and children. Therapeutic Drug Monitoring 13: 507-510, 1991 Bartosyk GD, Meyerson W, Reimann w, Satzinger G, Von Hodenborg A. Gabapentin. In Meldrum BS, Porter RJ (Eds) Current problems in epilepsy 4: new anticonvulsant drugs, pp. 147168, John Libby, London, 1986 Battino D, Bossi L, Croci D, Franceschetti S, Gomeni C, et al. Carbamazepine plasma levels in children and adults: influence of age, dose and associated therapy. Therapeutic Drug Monitoring 2: 315-322, 1980

C/in. Pharmacokinet. 23 (3) 1992

Beghi E, Di Mascio R, Tognoni G. Drug treatment of epilepsy: outlines, criticism and perspectives. Drugs 31: 249-265, 1986 Bertilsson L, Tomson T, Tybring G. Pharmacokinetics: time-dependent changes - autoinduction of carbamazepine epoxidation Journal of Clinical Pharmacology 26: 459-462, 1986 Bialer M. Pharmacokinetic evaluation of sustained release formulations of anti epileptic drugs. Clinical Pharmacokinetics 22: 11-21,1992 Binnie CD, Debets RMC, Engelsman M, Meijer JW A, Meinhardi H, et al. Double-blind crossover trial oflamotrigine (Lamictal) as add-on therapy in intractable epilepsy. Epilepsy Research 1989 4: 222-229, 1989 Binnie CD, Van Emde-Boas W, Kasteleijn-Nolste-Trenite DGA, De Korta RA, Meijer JW A, et al. Acute effects of lamotrigine (BW43OC) in persons with epilepsy. Epilepsia 27: 248-254, 1986 Blaser KU, Vozeh S, Landolt H, Kaufmann G, Romppainen J, et al. Intravenous phenytoin: a loading scheme for desired concentrations. Annals ofinternal Medicine 110: 1029-1031, 1989 Boucher BA, Rodman JH, Fabian TC, Cupit GC, Ludden TM, et al. Disposition of phenytoin in critically ill trauma patients. Clinical Pharmacy 6: 881-887, 1987 Boucher BA, Rodman JH, Jaresko GS, Rasmussen SN, Watridge CB, et al. Phenytoin pharmacokinetics in critically ill trauma patients. Clinical Pharmacology and Therapeutics 44: 675-683, 1988 Bourgeois BFD. Valproate: clinical use. In Levy R et al. (Eds) Antiepileptic drugs, 3rd ed., pp. 633-642, Raven Press, New York, 1989 Brodie MJ. Established anticonvulsants and treatment of refractory epilepsy. Lancet 336: 350-354, 1990a Brodie MJ. Management of epilepsy during pregnancy and lactation. Lancet 336: 426-427, 1990b Brodie MJ. Status epilepticus in adults. Lancet 336: 551-552, 1990c Brodie MJ. Drug interactions and epilepsy. Epilepsia, in press, I 992a Brodie MJ. Lamotrigine. Lancet, in press, 1992b Brodie MJ, Feely J. Practical clinical pharmacology. Therapeutic drug monitoring and clinical trials British Medical Journal 296: 1110-1114,1988 Brodie MJ, Forrest G, Rapeport WG. Carbamazepine 10,11epoxide concentrations in epileptics on carbamazepine alone and in combination with other anticonvulsants. British Journal of Clinical Pharmacology 16: 747-750, 1983 Brodie MJ, McPhail E, Macphee GJA, Larkin JG, Gray JMB. Psychomotor impairment and anticonvulsant therapy in adult epileptic patients. European Journal of Clinical Pharmacology 31: 655-660, 1987 Brodie MJ, Porter RJ. New and potential anticonvulsants. Lancet 336: 425-426, 1990 Browne T. The pharmacokinetics of agents used to treat status epilepticus. Neurology 40 (Suppl. 2): 29-42, 1990 Bryson SM, Al-Lanqawi Y, Kelman AW, Whiting B. Comparison of a Bayesian forecasting technique with a new method for estimating phenytoin dose requirements. Therapeutic Drug Monitoring 10: 80-84, 1988 Chadwick DW. Diagnosis of epilepsy. Lancet 336: 291-295, 1990a Chadwick D. Gabapentin in partial epilepsy. Lancet 335: 11141117, 1990a Chaikin P, Adir J. Unusual absorption profile of phenytoin in a massive overdose case. Journal of Clinical Pharmacology 27: 70-73, 1987 Choonara lA, Rane A. Therapeutic drug monitoring of anticonvulsants: state of the art. Clinical Pharmacokinetics 18: 318328, 1990 Chrystyn H, Morgan DH. A comparison of graphical nomogram methods with a computerised Bayesian analysis method in the interpretation of serum phenytoin concentrations. Journal of Clinical and Hospital Pharmacy II: 443-448, 1986 Crawford P, Ghadali E, Lane R, Bumhardt L, Chadwick D. Ga-

Anticonvulsant Optimisation

bapentin as an antiepileptic drug in man. Journal of Neurology, Neurosurgery and Psychiatry 50: 682-686, 1987 Crawford TO, Mitchell WG, Fishman LS, Snodgrass SR. Veryhigh-dose phenobarbital for refractory status epilepticus in children. Neurology 38: 1035-1040, 1988 Crowley JJ, Koup JR, Cusack BJ, Ludden TM, Vestal RE. Evaluation of a proposed method for phenytoin maintenance dose prediction following an intravenous loading dose. European Journal of Clinical Pharmacology 32: 141-148, 1987 Dam M, Ekberg R, Loyning Y, Waltimo 0, Jakobsen K. A doubleblind study comparing oxcarbazepine and carbamazepine in patients with newly diagnosed, previously untreated epilepsy. Epilepsy Research 3: 70-76, 1989 Duncan JS, Patsalos PN, Shorvon SD. Effects of discontinuation of phenytoin, carbamazepine and valproate on concomitant antiepileptic medication. Epilepsia 32: 101-115, 1991 Durelli L, Massazza U, Cavallo R. Carbamazepine toxicity and poisoning. Incidence, clinical features and management. Medical Toxicology and Adverse Drug Experience 4: 95-107, 1989 Eadie MJ. Plasma monitoring of anticonvulsants. Clinical Pharmacokinetics I: 52-66, 1976 Eadie MJ, Lander CM, Hooper WD, Tyrer JH. Factors influencing plasma phenobarbitone levels in epileptic patients. British Journal of Clinical Pharmacology 4: 541-547, 1977 Editorial. Sodium valproate. Lancet 2: 1229-1231, 1988 Editorial. Vigabatrin. Lancet I: 532-533, 1989a Editorial. Oxcarbazepine. Lancet 2: 196-198, 1989b Eichelbaum M, Tomson T, Tybring G, Bertilsson L. Carbamazepine metabolism in man. Induction and pharmacogenetic aspects. Clinical Pharmacokinetics 10: 80-90, 1985 Farrell K, Abbott FS, Orr JM, et al. Free and total valproate concentrations - their relationship to seizure control, liver enzymes and plasma ammonia in children. Canadian Journal of Neurological Science 13: 252-255, 1986 Flint N, Lopez LM, Robinson JD, Williams C, Salem RB. Comparison of eight phenytoin dosing methods in institutionalised patients. Therapeutic Drug Monitoring 7: 74-80, 1985 Gabor AJ. Lorazepam versus phenobarbital: candidate for drug of choice for treatment of status epilepticus. Journal of Epilepsy 3: 3-6, 1990 Garcia MJ, Alonso AC, Maza A, Santos D, Matesanz C, et al. Comparison of methods of carbamazepine dosage individualisation in epileptic patients. Journal of Clinical Pharmacy and Therapeutics 13: 375-380. 1988 Gillham RA, Williams M, Weidmann K. Butler E, Larkin JG, et al. Concentration-effect relationships with carbamazepine and its epoxide on psychomotor and cognitive function in epileptic patients. Journal of Neurology, Neurosurgery and Psychiatry 51: 929-933, 1988 Gillman ME, Toback JW, Gal P, Erikan NV. Individualising phenobarbital dosing in neonates. Clinical Pharmacy 2: 258262, 1983 Gillman JT, Gal P, Duchowny MS, Weaver RL, Ransom JL. Rapid sequential phenobarbital treatment of neonatal seizures. Paediatrics 83: 674-678, 1989 Godley PJ, Ludden TM, Clemanti WA, Godley SE, Ramsay RR. Evaluation of a Bayesian regression-analysis computer program using non-steady-state phenytoin concentrations. Clinical Pharmacy 6: 634-639, 1987 Grant SM, Heel Re. Vigabatrin. Drugs 41: 889-926, 1991 Graves NM, Leppik IE, Termond E, Taylor JW. Phenytoin clearances in a compliant population: description and application. Therapeutic Drug Moditoring 8: 427-433, 1986 Haegele KD, Huebert ND, Erel M, Tell GP, Schechter PJ. Pharmacokinetics of vigabatrin: implications of creatinine clearance. Clinical Pharmacology and Therapeutics 44: 558-585, 1988 Henriksen 0, Johannessen SI. Clinical and pharmacokinetic observations on sodium valproate - a 5-year follow-up study of

229

100 children with epilepsy. Acta Neurologica Scandinavica 65: 504-523, 1982 Howden CW, Birnie GG, Brodie MJ. Drug metabolism in liver disease. Pharmacology and Therapeutics 40: 439-474, 1989 Hudson SA, Farquhar DL, Thompson D, Smith RG. Phenytoin dosage individualisation - five methods compared in the elderly. Journal of Clinical Pharmacy and Therapeutics 15: 2534, 1990 Jawad S, Richens A, Goodwin G, Yuen AWe. Controlled trial ofiamotrigine for refractory partial seizures. Epilepsia 30: 356363, 1989 Jawad S, Yuen AWC, Peck AW, Hamilton MJ, Oxley JR, et aI. Lamotrigine: singl~ose pharmacokinetics and initial one-week experience in refractory seizures. Epilepsy Research I: 194-20 I, 1987 Johannessen SI. Antiepileptic drugs: pharmacokinetic and clinical aspects. Therapeutic Drug Monitoring 3: 17-37, 1981 Johnson FN, Brodie MJ. Pharmacokinetics of carbamazepine: implications for clinical use. Reviews in Contemporary Pharmacotherapy I: 85-98, 1990 Kelman AW, Whiting B, Bryson SM. OPT: a package of computer programs for parameter optimisation in clinical pharmacokinetics. British Journal of Clinical Pharmacology 14: 247256, 1982 Killilea T, Coleman R, Ludden T, Peck CC, Rose D. Bayesian regression analysis of non-steady-state phenytoin concentrations: evaluation of predictive performance. Therapeutic Drug Monitoring II: 455-462, 1989 Larkin JG, Herrick AL, McGuire GM, Percy-Robb I, Brodie MJ. Antiepileptic drug monitoring at the epilepsy clinic: a prospective evaluation. Epilepsia 32: 89-95, 1991 Larkin JG, McKee PJ, Brodie MJ. Rapid tolerance to acute psychomotor impairment with carbamazepine in epileptic patients. British Journal of Clinical Pharmacology 33: 1111-1114, 1992 Larkin JG, McKee PJ, Forrest G, Beastall GH, Park BK, et al. Lack of enzyme induction with oxcarbazepine (600mg daily) in healthy subjects. British Journal of Clinical Pharmacology 31: 65-71, 1991b Larkin JG, McLellan A, Munday A, Sutherland M, Butler E, et al. A double-blind comparison of conventional and controlledrelease carbamazepine in healthy subjects. British Journal of Clinical Pharmacology 27: 313-322, 1989 Levy RH, Kerr BM. Clinical pharmacokinetics of carbamazepine. Journal of Clinical Psychiatry 49 (Suppl. 4): 58-62, 1988 Loiseau P, Yuen AWC, Duche B, Menager T, Arne-Bes Me. A randomised double-blind, placebo-controlled, crossover addon trial ofiamotrigine in patients with treatment-resistant partial seizures. Epilepsy Research 7: 136-145, 1990 Ludden TM, Allen JP, Valutsky WA, Vicuna AV, Nappi JM, et al. Individualisation of phenytoin dosage regimens. Clinical Pharmacology and Therapeutics 21: 287-293, 1977 Ludden TM, Beal SL, Peck CC, Godley PJ. Evaluation ofa Bayesian regression-analysis program for predicting phenytoin concentration. Clinical Pharmacy 5: 580-585, 1986 Mcinnes GT, Brodie MJ. Drug interactions that matter - a critical reappraisal. Drugs 36: 83-110, 1988 McKee PJW, Blacklaw J, Butler E, Gillham RA, Brodie MJ. Monotherapy with conventional and controlled-release carbamazepine: a double-blind, double-