J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.48.11.1073 on 1 November 1985. Downloaded from Journal ofNeurology, Neurosurgery, and Psychiatry 1985;48: 1073-1077 Occasional review The comparative efficacy of antiepileptic drugs for partial and tonic-clonic seizures D CHADWICK, DM TURNBULL From the Mersey Regional Department ofMedical and Surgical Neurology, Walton Hospital, Liverpool, and the Department ofNeurology, University ofNewcastle-upon- Tyne, UK SUMMARY Studies of the efficacy of anticonvulsant drugs are difficult to undertake and histori- cally have been of poor quality. Randomised comparisons of drugs are few in number, and have failed to detect significant differences between drugs. This is surprising in view of the strong feelings that many clinicians have about the relative efficacy of the drugs they use. A review of the literature emphasises the need for further studies in this field. guest. Protected by copyright. Recent studies have emphasised that the majority of introduction of sodium valproate, drugs effective patients with the onset of epilepsy in adult life can against petit mal absence seizures in childhood (for satisfactorily be managed by using a single anticon- example ethosuximide and trimethadione), were vulsant drug.' There is little definite evidence that ineffective against tonic-clonic and partial seizures adding a second anticonvulsant drug improves seiz- whilst drugs such as phenytoin and carbamazepine ure control in patients who have received an optimal which are effective against the latter are ineffective dose of a single drug.24 Since the risks of adverse against petit mal absence seizures. However, it has reactions and chronic toxicity increase when also been asssumed that different drugs are to be multiple anticonvulsant drugs are administered,45 preferred depending on whether a patient has monotherapy has become increasingly popular. This tonic-clonic seizures, or a partial epilepsy. Porter6 poses a question that has not previously demanded has recently reviewed the subject and suggested that serious investigation, namely: which anticonvulsant there is an order of preference for drugs which drug to choose for an individual patient? depends on seizure type (table 1). This ranking of Three major factors are likely to influence the anticonvulsant efficacy conforms to generally choice of an anticonvulsant: efficacy, toxicity and accepted clinical practice and would not be regarded cost. In clinical practice assumptions made about as controversial by many neurologists. However, it is comparative efficacy of anticonvulsants have been of uncertain on what evidence these preferences are major importance in this choice, and it is the based as the literature relating to the efficacy of purpose of this review to question the basis for these anticonvulsant drugs is unsatisfactory.'-' Few http://jnnp.bmj.com/ assumptions. studies have included such basic requirements as a The desire to adopt a universally acceptable randomised design and adequate statistical analysis classification of both seizures and epilepsy stems in in comparing anticonvulsant drugs. These omissions part from the assumption that there is a different are compounded by two further problems: a lack of response of individual seizure types and epilepsies to a generally accepted definition for "control" of different anticonvulsant drugs. There is some epilepsy, and problems in the selection of patients. evidence to support this contention in that, until the Control of epilepsy on September 30, 2021 by Address for reprint requests: Dr D Chadwick, Walton Hospital, Rice Lane, Liverpool L9 1AE, UK. Response to an anticonvulsant drug is most com- monly described using arbitrary definitions of poor, Received 14 December 1984 and in revised form 14 February good or excellent control. These terms are usually 1985. Accepted 16 Februarv 1985 defined by comparing pre-test seizure frequency, 1073 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.48.11.1073 on 1 November 1985. Downloaded from 1 074 Chadwick, Turnbull Table 1 Orders ofeffectiveness against tonic-clonic and partial seizures Decreasing likelihood of Simple* partial seizures Complex* partial seizures Tonic-clonic seizures effectiveness 1 phenytoin carbamazepine phenytoin carbamazepine phenytoin carbamazepine 3 primidone primidone phenobarbitone 4 phenobarbitone phenobarbitone valproate (Porter, 1982)6 *as defined in 1981 International Classification of seizures. with seizure frequency following the initiation or drug is given as add-on therapy to patients with change of therapy, and using percentage change to chronic epilepsy, whose seizures continue in spite of define the response as "good", "poor" etc. one, two, or more drugs in combination, the lack of However, there may be inherent difficulties in such response to an additional drug may not imply a lack comparisons. Patients rarely have a stable seizure of anticonvulsant properties. It may simply be an threshold and seizures not infrequently occur in indication that the trial drug is no more effective clusters. Thus when a patient presents with a flurry than the previously administered drugs, or that the of four seizures within one week and has his or her trial drug has a similar mechanism of action to the treatment amended it might be assumed that the concurrently administered drugs. pre-treatment seizure frequency might amount to These considerations make it essential that com- some 16 seizures per month. However the patient parative studies of the efficacy and toxicity of anti- may only experience one such flurry of seizures in a convulsant drugs should rely on trial designs which year, in which case when treatment is started the include a randomisation procedure with prolonged apparent monthly seizure frequency might well fall follow-up. Multiple drug therapy should be avoided, guest. Protected by copyright. from twelve seizures per month, to 0-35 seizures per and the population of patients should be defined in month even if the drug treatment had no effect on terms of those factors known to influence prognosis. seizure frequency. It is therefore necessary in asses- The number of prospective randomised comparative sing treatment to ensure that the period of observa- studies has increased over the past decade but the tion prior to treatment is prolonged and identical to numbers still remain small, and many are still open the period following the change before comparisons to criticism. of seizure frequency can be derived. In previously untreated patients it may be difficult Randomised comparative studies to withhold a treatment to establish a pretreatment Two differing designs have been used. Most have seizure frequency so that other means of assessing employed a double blind cross-over design (tables 2 seizure response become necessary in such patients. and 3), which necessitates short periods of patient Here it seems preferable to adopt definitions of con- observation which may not provide information that trol of epilepsy which describe periods entirely free can readily be applied to longer-term clinical man- from seizures. In this instance control of epilepsy agement. Drugs such as benzodiazepines, whose may be defined in terms of the number of patients anticonvulsant effects show tolerance, are likely to entering remissions lasting, one, two or more be unduly favoured by such designs. Other studies years."' Such a definition is more clinically relevant have undertaken randomisation with a longer-term than any assessment based on comparisons of parallel-group follow-up (table 4). However, these seizure frequencies, clinicians and patients aiming usually give no information to confirm that the for complete cessation of attacks, rather than reduc- groups as randomised are comparable for those fac- http://jnnp.bmj.com/ tion of say 50% from four seizures per day, to two. tors known to influence the prognosis of epilepsy,'2 essential for this particular type of design, which Patient selection lacks a cross-over. In spite of these difficulties it is Studies of anticonvulsant efficacy have usually been nevertheless quite striking that only one study'" sug- undertaken in chronic epileptic patients who have gests that one of the anticonvulsant drugs tested was proved themselves resistant to therapy over many superior to another, phenytoin being preferred to years, and who often continue with their previous sulthiame. As some of the antiepileptic properties of medication during the drug study. The long-term sulthiame may be due to its ability to inhibit pheny- on September 30, 2021 by remission rates in such patients are always likely to toin metabolism,'4 it is perhaps not surprising that remain low, in contrast to the prognosis for newly the drug is inferior when compared to phenytoin. diagnosed patients with epilepsy in whom the prog- Three recently reported studies demand further nosis is considerably better.'°0 Indeed when a trial comment, as they were undertaken in previously J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.48.11.1073 on 1 November 1985. Downloaded from The comparative efficacy of antiepileptic drugs for partial and tonic-clonic seizures 1075 untreated patients who were randomised to differ- meaningful differences in efficacy in a parallel group ent drugs shortly after the diagnosis of epilepsy. Cal- design. laghan et all- compared phenytoin, carbamazepine, Why, then, should this array of data fail to dif- and valproate in 181 adult patients with tonic-clonic