Phenytoin versus valproate monotherapy for partial onset seizures and generalised onset tonic-clonic seizures

  • Review
  • Intervention

Authors


Abstract

Background

This is an updated version of the previously published Cochrane review (Issue 4, 2009)

Worldwide, phenytoin and valproate are commonly used antiepileptic drugs. It is generally believed that phenytoin is more effective for partial onset seizures, and that valproate is more effective for generalised onset tonic-clonic seizures with or without other generalised seizure types.

Objectives

To review the best evidence comparing phenytoin and valproate when used as monotherapy in individuals with partial onset seizures or generalised onset tonic-clonic seizures with or without other generalised seizure types.

Search methods

We searched the Cochrane Epilepsy Group's Specialised Register (19 February 2013), the Cochrane Central Register of Controlled Trials (CENTRAL, Issue 1, The Cochrane Library, January 2013), MEDLINE (1946 to 18 February 2013), SCOPUS (19 February 2013), ClinicalTrials.gov (19 February 2013), and WHO International Clinical Trials Registry Platform ICTRP (19 February 2013). We handsearched relevant journals, contacted pharmaceutical companies, original trial investigators and experts in the field.

Selection criteria

Randomised controlled trials in children or adults with partial onset seizures or generalised onset tonic-clonic seizures with a comparison of valproate monotherapy versus phenytoin monotherapy.

Data collection and analysis

This was an individual patient data review. Outcomes were time to (a) treatment withdrawal (b) 12-month remission (c) six-month remission and (d) first seizure post randomisation. Cox proportional hazards regression models were used to obtain study-specific estimates of hazard ratios (HRs) with 95% confidence intervals (CIs) with the generic inverse variance method used to obtain the overall pooled HR and 95% CI.

Main results

Individual patient data were available for 669 individuals out of 1119 eligible individuals from five out of 11 trials, 60% of the potential data. Results apply to generalised tonic-clonic seizures, but not absence or myoclonus seizure types. For remission outcomes, HR > 1 indicates an advantage for phenytoin and for first seizure and withdrawal outcomes HR > 1 indicates an advantage for valproate

The main overall results (pooled HR adjusted for seizure type, 95% CI) were time to (a) withdrawal of allocated treatment 1.09 (0.76 to 1.55); (b) 12-month remission 0.98 (0.78 to 1.23); (c) six-month remission 0.95 (0.78 to 1.15) and (d) first seizure 0.93 (0.75 to 1.14). The results suggest no overall difference between the drugs for these outcomes. No statistical interaction between treatment and seizure type (partial versus generalised) was found, but misclassification of seizure type may have confounded the results of this review.

Authors' conclusions

We have not found evidence that a significant difference exists between phenytoin and valproate for the outcomes examined in this review. However misclassification of seizure type may have confounded the results of this review. Results do not apply to absence or myoclonus seizure types. No outright evidence was found to support or refute current treatment policies.

Résumé scientifique

Comparaison entre la phénytoïne et le valproate en monothérapie pour traiter les crises d'épilepsie partielle et les crises toniques-cloniques généralisées

Contexte

Ceci est une version mise à jour de la revue Cochrane publiée dans le numéro 4 en 2009.

Dans le monde entier, la phénytoïne et le valproate sont des antiépileptiques couramment utilisés. On considère généralement que la phénytoïne est plus efficace pour les crises d'épilepsie partielle, et que le valproate est plus efficace pour les crises toniques-cloniques généralisées avec ou sans autres types de crises généralisées.

Objectifs

Évaluer les meilleures données comparant la phénytoïne et le valproate lorsqu'ils sont utilisés en monothérapie chez les participants souffrant de crises d'épilepsie partielle ou de crises toniques-cloniques généralisées, avec ou sans autres types de crises généralisées.

Stratégie de recherche documentaire

Nous avons effectué une recherche dans le registre spécialisé du groupe Cochrane sur l'épilepsie (mardi 19 février 2013), le registre Cochrane des essais contrôlés (CENTRAL, numéro 1, The Cochrane Library janvier 2013), MEDLINE (de 1946 au 18 février 2013), SCOPUS (19.02.13), ClinicalTrials.gov (19.02.13), et le système d'enregistrement international des essais cliniques de l'OMS (ICTRP) (19.02.13). Nous avons effectué des recherches manuelles dans des journaux pertinents, contacté des laboratoires pharmaceutiques, les investigateurs des essais originaux et des experts du domaine.

Critères de sélection

Essais contrôlés randomisés chez les enfants ou les adultes souffrant de crises d'épilepsie partielle ou de crises toniques-cloniques généralisées utilisant une comparaison entre la monothérapie au valproate et la monothérapie à la phénytoïne.

Recueil et analyse des données

Il s'agissait d'une revue portant sur des données individuelles des participants (DIP). Les critères étaient le temps jusqu'à (a) l'arrêt du traitement (b) une rémission de 12 mois (c) une rémission de six mois et (d) la première crise après la randomisation. Des modèles de régression à risques proportionnels de Cox ont été utilisés pour obtenir les estimations spécifiques aux études des hazard ratios (HR) avec des intervalles de confiance (IC) à 95 %, la méthode générique de l'inverse de la variance étant utilisée pour obtenir l'estimation regroupée globale des HR et des IC à 95 %.

Résultats principaux

Les données individuelles des participants étaient disponibles pour 669 patients sur les 1 119 personnes éligibles dans cinq essais sur les onze, soit 60 % des données potentielles. Les résultats s'appliquent aux crises toniques-cloniques généralisées, mais pas à l'absence de crise ou aux types de crises myocloniques. Pour les résultats de rémission, HR > 1 indique un avantage pour la phénytoïne, et pour les résultats de délai jusqu'à la première crise et jusqu'à l'arrêt du traitement, HR > 1 indique un avantage pour le valproate

Les principaux résultats globaux (HR regroupé ajusté pour le type de crise, IC à 95 %) étaient (a) le temps jusqu'à l'arrêt du traitement attribué 1,09 (0,76 à 1,55) ; (b) le temps jusqu'à l'obtention d'une rémission à 12 mois 0,98 (0,78 à 1,23) ; (c) le temps jusqu'à l'obtention d'une rémission à six mois 0,95 (0,78 à 1,15) et (d) le temps jusqu'à la première crise 0,93 (0,75 à 1,14). Les résultats suggèrent qu'il n'y a pas de différence globale entre les médicaments pour ces critères. Aucune interaction statistique n'a été détectée entre le traitement et le type de crises (partielle comparée à généralisée), mais des erreurs de classification dans le type de crise ont pu fausser les résultats de cette revue.

Conclusions des auteurs

Nous n'avons trouvé aucune preuve indiquant qu'il existe une différence significative entre la phénytoïne et le valproate pour les critères examinés dans cette revue. Toutefois, l'erreur de classification du type de crise a pu fausser les résultats de cette revue. Les résultats ne s'appliquent pas à l'absence de crise ou aux types de crises myocloniques. Aucune preuve irréfutable n'a été identifiée pour confirmer ou récuser les stratégies thérapeutiques actuelles.

Plain language summary

Phenytoin versus valproate monotherapy for partial onset seizures and generalised onset tonic-clonic seizures

Epilepsy is a disorder in which recurrent seizures are caused by abnormal electrical discharges from the brain. We studied two seizure types in this review; generalised onset seizures in which electrical discharges begin in one part of the brain and move throughout the brain, and partial onset seizures in which the seizure is generated in and affects one part of the brain (the whole hemisphere of the brain or part of a lobe of the brain). Most seizures can be controlled by a single antiepileptic drug. Worldwide, phenytoin and valproate are commonly used antiepileptic drugs. This review of trials found no difference between these two drugs for the seizure types studied.The review also found no evidence to support or refute the policy of using valproate for generalised onset tonic-clonic seizures and phenytoin for partial onset seizures. However, up to 49% of people within the trials classified as having generalised seizures may have had their seizure type wrongly diagnosed, and this misclassification may have influenced the results of this review. We were unable to address the issue of preferring valproate for generalised onset seizure types other than tonic-clonic.

Résumé simplifié

Comparaison entre la phénytoïne et le valproate en monothérapie pour traiter les crises d'épilepsie partielle et les crises toniques-cloniques généralisées

L'épilepsie est un trouble qui se caractérise par des convulsions récurrentes causées par des décharges électriques anormales dans le cerveau. Nous avons étudié deux types de crises dans cette revue : les crises d'épilepsie généralisées dans lesquelles les décharges électriques commencent dans une partie du cerveau et se déplacent dans l'ensemble du cerveau, et les crises d'épilepsie partielle dans lesquelles la crise est générée dans et affecte une partie du cerveau (l'hémisphère entier du cerveau ou une partie d'un lobe du cerveau). La plupart des crises peuvent être maîtrisées par un médicament antiépileptique unique. Dans le monde entier, la phénytoïne et le valproate sont des antiépileptiques couramment utilisés. Cette revue d'essais n'a trouvé aucune différence entre ces deux médicaments pour les types de crises étudiés.De même, la revue n'a pas trouvé de preuves pour confirmer ou récuser la stratégie d'utilisation du valproate pour les crises toniques-cloniques généralisées et de la phénytoïne pour les crises d'épilepsie partielle. Toutefois, il est possible que le type de crise diagnostiqué ait été incorrect chez jusqu'à 49 % des personnes incluses dans les essais et classées dans la catégorie des crises généralisées, et que cette erreur de classification puisse avoir influencé les résultats de cette revue. Nous n'avons pas été en mesure de trancher la question de la préférence du valproate pour les crises généralisées autres que les crises toniques-cloniques.

Notes de traduction

Traduit par: French Cochrane Centre 16th October, 2013
Traduction financée par: Pour la France : Minist�re de la Sant�. Pour le Canada : Instituts de recherche en sant� du Canada, minist�re de la Sant� du Qu�bec, Fonds de recherche de Qu�bec-Sant� et Institut national d'excellence en sant� et en services sociaux.

Summary of findings(Explanation)

Summary of findings for the main comparison. 
  1. 1 Pooled Hazard Ratio for all participants adjusted for seizure type

    2 Risk of Bias judged high for four unblinded studies (Craig 1994; De Silva 1996; Heller 1995; Ramsay 1992)

    3 Up to 190 out of 384 (49%) of adult participants (in Craig 1994; De Silva 1996; Heller 1995; Ramsay 1992; Shakir 1981; Turnbull 1985) may have had their seizure type wrongly classified as generalised onset; sensitivity analyses show misclassification has an impact on results and conclusions

    4 Sensitivity analysis for misclassification of epilepsy type shows similar results and unchanged conclusions, so misclassification is unlikely to impact on results

    5 Only one trial (Ramsay 1992) collected data on generalised seizure types other than generalised tonic-clonic seizures. Hence, the results apply only to generalised tonic-clonic seizures, despite the fact that individuals may have been experiencing other generalised seizure types.

Phenytoin compared with valproate for partial onset seizures and generalised onset tonic-clonic seizures

Patient or population: adults and children with newly-onset partial or generalised tonic-clonic seizures

Settings: Outpatients

Intervention: Valproate

Comparison: Phenytoin

OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
PhenytoinValproate

Time to withdrawal of allocated treatment

- stratified by epilepsy type

Range of follow-up (all participants): 1 - 91 months

27 per 100

25 per 100

(18 to 33)

HR 1.09

(0.76 to 1.55)1

528
(5 studies)
⊕⊕⊕⊝
moderate 2,4
HR > 1 indicates a clinical
advantage for valproate

Time to withdrawal of allocated treatment

- stratified by epilepsy type

- Generalised onset seizures (tonic-clonic only)

Range of follow-up (all participants): 1 - 91 months

18 per 100

19 per 100

(12 to 29)

HR 0.98

(0.59 to 1.64)

341
(5 studies)
⊕⊕⊕⊝
moderate 2,4
HR > 1 indicates a clinical
advantage for valproate

Time to withdrawal of allocated treatment

- stratified by epilepsy type - Partial onset seizures

Range of follow-up (all participants): 1 - 91 months

39 per 100

34 per 100

(23 to 49)

HR 1.20

(0.74 to 1.95)

187

(4 studies)

⊕⊕⊕⊝
moderate2,4
HR > 1 indicates a clinical
advantage for valproate

Time to 12-month remission - stratified by epilepsy type

Range of follow-up (all participants): 1 - 91 months

67 per 100

67 per 100

(58 to 75)

HR 0.98

(0.78 to 1.23)1

514

(4 studies)

⊕⊕⊕⊝
moderate 2,4
HR > 1 indicates a clinical
advantage for phenytoin

Time to 12-month remission - stratified by epilepsy type

- Generalized onset seizures (tonic-clonic only)

Range of follow-up (all participants): 1 - 91 months

67 per 100

69 per 100

(58 to 79)

HR 1.04

(0.77 to 1.40)

270

(4 studies)

⊕⊕⊝⊝
low 2,4,5
HR > 1 indicates a clinical
advantage for phenytoin

Time to 12-month remission - stratified by epilepsy type

- Partial onset seizures

Range of follow up (all participants): 1 - 91 months

67 per 100

63 per 100

(50 to 76)

HR 0.90

(0.63 to 1.29)

244

(4 studies)

⊕⊕⊕⊝
moderate2,4
HR > 1 indicates a clinical
advantage for phenytoin

Time to 6-month remission - stratified by epilepsy type

Range of follow-up (all participants): 1 - 91 months

60 per 100

58 per 100

(51 to 65)

HR 0.95

(0.78 to 1.15)1

639

(5 studies)

⊕⊕⊕⊝
moderate 2,4
HR > 1 indicates a clinical
advantage for phenytoin

Time to 6-month remission - stratified by epilepsy type

- Generalised onset seizures (tonic-clonic only)

Range of follow-up (all participants): 1 - 91 months

69 per 100

66 per 100

(57 to 75)

HR 0.92

(0.72 to 1.18)

395

(5 studies)

⊕⊕⊝⊝
low 2,4,5
HR > 1 indicates a clinical
advantage for phenytoin

Time to 6-month remission - stratified by epilepsy type

- Partial onset seizures

Range of follow-up (all participants): 1 91 months

51 per 100

50 per 100

(40 to 62)

HR 0.99

(0.73 to 1.35)

244

(4 studies)

⊕⊕⊕⊝
moderate2,4
HR > 1 indicates a clinical
advantage for phenytoin

Time to first seizure - stratified by epilepsy type

Range of follow-up (all participants): 1 - 91 months

59 per 100

62 per 100

(55 to 70)

HR 0.93

(0.75 to 1.14)1

639

(5 studies)

⊕⊕⊝⊝
low2,3
HR>1 indicates a clinical
advantage for valproate

Time to first seizure - stratified by epilepsy type

- Generalised onset seizures (tonic-clonic only)

Range of follow-up (all participants): 1 91 months

48 per 100

47 per 100

(38 to 58)

HR 1.03

(0.77 to 1.39)

395

(5 studies)

⊕⊝⊝⊝
very low 2,3,5
HR > 1 indicates a clinical
advantage for valproate

Time to first seizure - stratified by epilepsy type

- Partial onset seizures

Range of follow-up (all participants): 1 91 months

75 per 100

81 per 100

(71 to 89)

HR 0.83

(0.62 to 1.11)

244

(4 studies)

⊕⊕⊝⊝
low 2,3
HR > 1 indicates a clinical
advantage for valproate

*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes.The assumed risk is calculated as the event rate in the Phenytoin treatment group

The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
The corresponding risk is calculated as the assumed risk x the relative risk (RR) of the intervention where RR = (1 - exp(HR x ln(1 - assumed risk)) ) / assumed risk
CI: Confidence interval; RR: relative risk; HR: Hazard Ratio; exp: exponential

GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Background

This review is an update of a previously published review in the Cochrane Database of Systematic Reviews (Issue 4, 2009).

The majority of people with epilepsy have their seizures controlled by a single drug (monotherapy) (Cockerell 1995).Worldwide, phenytoin and valproate are commonly used antiepileptic drugs (AEDs) licensed for monotherapy. Phenytoin is used as a first-line drug in low- and middle-income countries as it is cheaper and can be given as a single daily dose, but is no longer considered a first-line agent in the USA and much of Europe due to worries over adverse events (Wallace 1997; Wilder 1995). Phenytoin is associated with long-term cosmetic changes including gum hyperplasia, acne and coarsening of the facial features (Mattson 1985; Scheinfeld 2003), as well as low folic acid levels, predisposing participants to megaloblastic anaemia (Carl 1992), and is associated with congenital abnormalities (Gladstone 1992; Morrow 2006; Meador 2008; Nulman 1997), particularly foetal hydantoin syndrome (Scheinfeld 2003). Furthermore, due to the pharmacokinetic profile of phenytoin, the plasma concentrations are difficult to predict and dosing will usually need to be informed by measuring plasma concentration. Valproate has also been shown to have teratogenic properties (Canger 1999; Morrow 2006; Tomson 2011) and is particularly associated with spina bifida and cardiac, craniofacial, skeletal and limb defects known as 'valproate syndrome' (Ornoy 2009). A systematic review found valproate to have the highest incidence of congenital malformations of standard AEDs (Meador 2008), and a recent study has shown an increased prevalence of neurodevelopmental disorders following prenatal valproate exposure (Bromley 2013). Valproate is also associated with weight gain in adults (Dinesen 1984; Easter 1997) and children (Egger 1981; Novak 1999).

It is generally believed that valproate monotherapy is more effective than phenytoin monotherapy in generalised onset seizures (generalised tonic-clonic seizures, absence and myoclonus), while phenytoin monotherapy is more effective than valproate monotherapy in partial onset seizures (simple partial, complex partial and secondary generalised tonic-clonic seizures) (Chadwick 1994), although there is no conclusive evidence from individual randomised controlled trials to support this belief. Evidence in favour of valproate for generalised seizures is predominantly anecdotal from observational studies, suggesting a dramatic benefit with valproate in juvenile myoclonic epilepsy (Delgado-Escueta 1984; Penry 1989), and reports of efficacy of valproate against absence seizures (Bourgeois 1987; Jeavons 1977). The results of two randomised controlled trials recruiting children (De Silva 1996; Thilothammal 1996) indicate that valproate may be better tolerated in children than phenytoin; twice as many children experienced at least one side effect on phenytoin than valproate in Thilothammal 1996 and phenytoin was more likely to be withdrawn due to unacceptable side effects than valproate in De Silva 1996.

Some animal models have suggested that phenytoin has either no effect in absence seizures or may in fact worsen seizures (Liporace 1994). There is also anecdotal evidence that phenytoin may cause paradoxical intoxication (increased seizure frequency with increased anticonvulsant dose) and encephalopathy (Troupin 1975; Vallarta 1974). Accepting that phenytoin should not be a drug of first choice for individuals experiencing absence, myoclonic and atonic seizures, we still have insufficient evidence from randomised controlled trials to guide a choice between phenytoin and valproate for individuals with generalised onset tonic-clonic seizures or partial onset seizures. The aim of this review, therefore, is to summarise efficacy and tolerability data from existing trials comparing phenytoin and valproate when used as monotherapy treatments.

There are difficulties in undertaking a systematic review of epilepsy monotherapy trials as the important efficacy outcomes require analysis of time-to-event data (for example, time to first seizure after randomisation). Although methods have been developed to synthesise time-to-event data using summary information (Parmar 1998; Williamson 2002), the appropriate statistics are not commonly reported in published epilepsy trials.

Furthermore, although seizure data have been collected in most epilepsy monotherapy trials, there has been no uniformity in the definition and reporting of outcomes. For example, trials may report time to 12-month remission but not time to first seizure or vice versa, or some trials may define time to first seizure from the date of randomisation while others use date of achieving maintenance dose. Trial investigators have also adopted differing approaches to the analysis, particularly with respect to the censoring of time-to-event data. For these reasons, we performed this review using individual patient data (IPD) which helps to overcome these problems. This review is one in a series of Cochrane IPD reviews investigating pair wise monotherapy comparisons. These data have also been included in a network meta-analysis (Tudur Smith 2007), undertaken following a previous version of this review.

Objectives

To review the time to withdrawal, remission and first seizure of phenytoin compared to valproate when used as monotherapy in people with partial onset seizures or generalised tonic-clonic seizures with or without other generalised seizure types.

Methods

Criteria for considering studies for this review

Types of studies

(1) Randomised controlled trials (RCTs) using either:

  • an adequate method of allocation concealment (e.g. sealed opaque envelopes);

  • a 'quasi' method of randomisation (e.g. allocation by date of birth).

(2) Studies may be double-blind, single-blind or unblinded.

(3) Studies must include a comparison of phenytoin monotherapy with valproate monotherapy in individuals with epilepsy.

Types of participants

(1) Children or adults with partial onset seizures (simple partial, complex partial, or secondarily generalised tonic-clonic seizures) or generalised onset tonic-clonic seizures (with or without other generalised seizure types).
(2) Individuals with a new diagnosis of epilepsy, or who have had a relapse following antiepileptic monotherapy withdrawal.

Types of interventions

Phenytoin (PHT) or valproate (SV) as monotherapy.

Types of outcome measures

Below is a list of outcomes investigated in this review. Reporting of these outcomes in the original trial report was not an eligibility requirement for inclusion in this review.

Primary outcomes

Time to withdrawal of allocated treatment (retention time). This is a combined outcome reflecting both efficacy and tolerability as treatment may be withdrawn due to continued seizures, adverse events or a combination of both. This is an outcome to which the participant makes a contribution, and is the primary outcome measure recommended by the Commission on Antiepileptic Drugs of the International League Against Epilepsy (Commission 1998; ILAE 2006)

Secondary outcomes

(1) Time to achieve 12-month seizure-free period (remission).
(2) Time to achieve six-month seizure-free period (remission).
(3) Time to first seizure post randomisation.

Search methods for identification of studies

Electronic searches

We searched the following databases. There were no language restrictions.

(a) The Cochrane Epilepsy Group's Specialised Register (19 February 2013) using the search strategy outlined in Appendix 1.
(b) The Cochrane Central Register of Controlled Trials (CENTRAL) (Issue 1,The Cochrane Library January 2013) using the search strategy outlined in Appendix 2.
(c) MEDLINE (Ovid, 1946 to 18 February 2013) using the search strategy outlined in Appendix 3.
(d) SCOPUS (19 February 2013) using the search strategy outlined in Appendix 4.
(e) ClinicalTrials.gov (19 February 2013) using the search terms 'phenytoin and valproate and epilepsy'.
(f) WHO International Clinical Trials Registry Platform ICTRP (19 February 2013) using the search terms 'phenytoin and valproate and epilepsy'.

Searching other resources

In addition, we handsearched relevant journals, reviewed the reference lists of retrieved studies to search for additional reports of relevant studies, contacted Sanofi (manufacturers of valproate in Europe), Abbott (manufacturers of valproate in the USA), Parke-Davis (manufacturers of phenytoin) and experts in the field for information about any ongoing studies.

Data collection and analysis

Selection of studies

Two review authors (SJN and AGM) independently assessed trials for inclusion, resolving any disagreements by mutual discussion.

Data extraction and management

The following individual patient data were requested for all trials meeting our inclusion criteria.

(1) Trial methods:

  • method of generation of random list;

  • method of concealment of randomisation;

  • stratification factors;

  • blinding methods.

(2) Participant covariates:

  • gender;

  • age;

  • seizure types;

  • time between first seizure and randomisation;

  • number of seizures prior to randomisation (with dates);

  • presence of neurological signs;

  • Electroencephalographic (EEG) results;

  • Computerised tomography/magnetic resonance imaging (CT/MRI) results.

(3) Follow-up data:

  • treatment allocation;

  • date of randomisation;

  • dates of follow-up;

  • dates of seizures post randomisation or seizure frequency data between follow-up visits;

  • dates of treatment withdrawal and reasons for treatment withdrawal;

  • dose;

  • dates of dose changes.

For each trial for which individual patient data (IPD) were not obtained, we carried out an assessment to see whether any relevant aggregate level data had been reported.

  • For three trials (Craig 1994; Ramsay 1992; Turnbull 1985) seizure data were provided in terms of the number of seizures recorded between clinic visits rather than specific dates of seizures. To enable time-to-event outcomes to be calculated, we applied linear interpolation to approximate the dates on which seizures occurred. For example, if four seizures were recorded between two visits which occurred on March 1st and May 1st (an interval of 61 days), then date of first seizure would be approximately March 13th. This allowed an estimate of the time to six-month and 12-month remission and the time to first seizure to be computed.

  • We calculated time to six-month and 12-month remission from the date of randomisation to the date (or estimated date) the individual had first been free of seizures for six or 12 months respectively. If the person had one or more seizures in the titration period, a six-month or 12-month seizure-free period could also occur between the estimated date of the last seizure in the titration period and the estimated date of the first seizure in the maintenance period.

  • We calculated time to first seizure from the date of randomisation to the date that their first seizure was estimated to have occurred. If seizure data were missing for a particular visit, these outcomes were censored at the previous visit. These outcomes were also censored if the individual died or if follow-up ceased prior to the occurrence of the event of interest. These methods had been used in the remaining two trials (De Silva 1996; Heller 1995) for which outcome data were provided directly.

  • Withdrawal data were not available for one trial (Craig 1994). For two trials (De Silva 1996; Heller 1995) we extracted dates and reason for treatment withdrawal from trial case report forms for the original review. Two review authors independently extracted data from all case report forms, resolving disagreements by reconsidering the case report forms at conference. For the remaining trials, data on length of time spent in trial and reason for withdrawal of allocated treatment were provided directly. For the analysis of time to event, an 'event' was defined as either the withdrawal of the allocated treatment due to poor seizure control or adverse events or both. Non-compliance with the treatment regimen or the addition of another antiepileptic drug were also classed as 'events'. The outcome was censored if treatment was withdrawn because the individual achieved a period of remission or if the individual was still on allocated treatment at the end of follow-up

Assessment of risk of bias in included studies

Two review authors (SJN and JP) independently assessed all included studies for risk of bias , resolving any disagreements by discussion.

Measures of treatment effect

All outcomes in this review were measured as time-to-event outcomes with the hazard ratio used as the measure of treatment effect. We calculated outcomes from IPD provided where possible, or extracted from published studies.

Unit of analysis issues

We did not have any unit of analysis issues. The unit of allocation and analysis was individual for all included studies and no studies were of a repeated measures (longitudinal) nature or of a crossover design.

Dealing with missing data

For each trial where IPD were supplied, results were reproduced from trial results where possible and consistency checks were performed:
(a) We cross-checked trial details against any published report of the trial and contacted original trial authors if we found missing data, errors or inconsistencies.
(b) We reviewed the chronological randomisation sequence, and checked the balance of prognostic factors, taking account of factors stratified for in the randomisation procedure.

Assessment of heterogeneity

We assessed heterogeneity statistically using the Q test (P value < 0.10 for significance) and the I² statistic (Higgins 2003) (greater than 50% indicating considerable heterogeneity), output produced using the generic inverse variance approach in Metaview, and visually by inspecting forest plots.

Assessment of reporting biases

Two review authors (SJN and JP) undertook all full quality and risk of bias assessments. In theory, a review using IPD should overcome issues of reporting biases as unpublished data can be provided and unpublished outcomes calculated. Any selective reporting bias detected could be assessed with the ORBIT classification system (Kirkham 2010).

Data synthesis

We carried out our analysis on an intention-to-treat basis (that is, participants were analysed in the group to which they were randomised, irrespective of which treatment they actually received). Therefore for the time-to-event outcomes 'Time to six-month remission', 'Time to 12-month remission' and 'Time to first seizure post randomisation' participants were not censored if treatment was withdrawn.

For all outcomes, we investigated the relationship between the time-to-event and treatment effect of the anti-epileptic drugs. We used Cox proportional hazards regression models to obtain study-specific estimates of log(hazard ratio) or treatment effect and associated standard errors in statistical software SAS version 9.2. (Copyright, SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA.). The model assumes that the ratio of hazards (risks) between the two treatment groups is constant over time (i.e. hazards are proportional). This proportional hazards assumption of the Cox regression model was tested for each outcome of each study by testing the statistical significance of a time-varying covariate in the model. We evaluated overall estimates of hazard ratios (with 95% confidence intervals) using the generic inverse variance method in Metaview.

Results are expressed as a hazard ratio (HR) and a 95% confidence interval (CI). By convention a HR greater than 1 indicates that an event is more likely to occur earlier on phenytoin than on valproate. Hence, for time to withdrawal of allocated treatment or time to first seizure, a HR greater than 1 indicates a clinical advantage for valproate (e.g. HR = 1.2 would suggest a 20% increase in risk of withdrawal from phenytoin compared to valproate) and for time to six-month and 12-month remission a HR greater than 1 indicates a clinical advantage for phenytoin.

Subgroup analysis and investigation of heterogeneity

Due to the strong clinical belief that valproate is more effective in generalised onset seizures while phenytoin is more effective in partial onset seizures, we have stratified all analyses by seizure type (partial onset versus generalised onset), according to the classification of main seizure type at baseline. Partial seizures (simple or complex) and partial secondarily generalised seizures were classified as partial epilepsy. Primarily generalised seizures were classified as generalised epilepsy. We conducted a Chi² test of interaction between treatment and epilepsy type.

If we found significant statistical heterogeneity to be present, we performed meta-analysis with a random-effects model in addition to a fixed-effect model, presenting the result of both models and performing sensitivity analyses to investigate differences in study characteristics.

Sensitivity analysis

As misclassification of seizure type is a recognised problem in epilepsy (whereby some individuals with generalised seizures have been mistakenly classed as having partial onset seizures and vice versa), we investigated its potential impact on results in a sensitivity analysis. Given clinical evidence that individuals with generalised onset seizures are unlikely to have an ' age of onset' greater than 25 to 30 years (Malafosse 1994), we examined the distribution of age at onset for individuals with generalised seizures. We undertook two sensitivity analyses to investigate misclassification:
(i) We reclassified all individuals with generalised seizure types and age at onset greater than 30 into an 'uncertain seizure type' group;
(ii) We reclassified individuals with generalised seizures and age of onset greater than 30 as having partial onset seizures.

Results

Description of studies

Results of the search

We identified 334 records from the databases and search strategies outlined in Electronic searches. We found no further records by searching other resources. We removed 126 duplicate records and screened 208 records (title and abstract) for inclusion in the review. We excluded 178 records based on title and abstract and assessed 30 full-text articles for inclusion in the review. We excluded 19 studies from the review (see Excluded studies below) and included 11 trials in the review (see Included studies below). See Figure 1 for PRISMA study flow diagram

Figure 1.

Study flow diagram.

Included studies

We included 11 trials in the review (Callaghan 1985; Czapinski 1997a; Craig 1994; De Silva 1996; Forsythe 1991; Heller 1995; Ramsay 1992; Rastogi 1991; Shakir 1981; Thilothammal 1996; Turnbull 1985). One trial was available abstract form only (Czapinski 1997a).

Four trials recruited individuals of all ages (Callaghan 1985; Ramsay 1992; Rastogi 1991; Shakir 1981), three trials recruited adults only (Czapinski 1997a; Heller 1995;Turnbull 1985), three trials recruited children only (De Silva 1996; Forsythe 1991; Thilothammal 1996) and one trial recruited elderly individuals only (Craig 1994). One trial recruited individuals with partial onset seizures only (Czapinski 1997a), two trials recruited individuals with generalised onset seizures only (Ramsay 1992; Thilothammal 1996), seven trials recruited individuals with partial onset seizures and generalised onset seizures (Callaghan 1985; Craig 1994; De Silva 1996; Heller 1995; Rastogi 1991; Shakir 1981; Turnbull 1985) and one trial did not provide information on the seizure types of individuals recruited (Forsythe 1991). Nine trials recruited individuals with new onset seizures only (Callaghan 1985; Craig 1994; Czapinski 1997a; De Silva 1996; Forsythe 1991; Heller 1995; Ramsay 1992; Thilothammal 1996; Turnbull 1985); 64% of individuals in one trial had new onset seizures, while the remaining individuals had uncontrolled seizures on current therapy (Shakir 1981); and one trial did not specify whether individuals were newly diagnosed (Rastogi 1991). Seven trials were conducted in Europe (Callaghan 1985; Craig 1994; Czapinski 1997a; De Silva 1996; Forsythe 1991; Heller 1995; Turnbull 1985),one trial in the USA (Ramsay 1992), two trials in India (Rastogi 1991; Thilothammal 1996) and one trial in two centres in Europe and New Zealand (Shakir 1981).

Individual patient data (IPD) were could not be provided for six of these trials (Callaghan 1985;Czapinski 1997a; Forsythe 1991; Rastogi 1991; Shakir 1981; Thilothammal 1996) in which a total of 450 individuals had been randomised to either phenytoin or valproate. None of these six trials reported the specific time-to-event outcomes chosen for this systematic review.

Two trials (Forsythe 1991; Shakir 1981) presented times at which allocated drug was withdrawn and the reason for withdrawal in the trial publication for each individual. Hence, these two trials could be incorporated into the analysis of 'Time to withdrawal of allocated treatment'; one of the trials (Shakir 1981) also presented information by seizure type (partial onset or generalised onset seizures) and therefore could also be included in the stratified analysis for 'Time to withdrawal of allocated treatment'. Shakir 1981 presents 'Time on trial drug' in months for each participant; therefore to calculate 'Time to withdrawal of allocated treatment' we assumed that if 'Time spent on trial drug' was five months, the individual spent five full months (152 full days) on the trial drug before withdrawal. Forsythe 1991 presents 'Withdrawal and time of occurrence by month' for each participant; therefore to calculate 'Time to withdrawal of allocated treatment' we assumed that if withdrawal occurred during the fifth month, that withdrawal occurred halfway between the fifth and sixth month (i.e. participants spent 167 full days on treatment before withdrawal).

We could not extract sufficient aggregate data from the trial publication in any other trial, and we therefore could not include them in data synthesis. Full details of outcomes considered and a summary of results of each trial for which IPD were not available to us can be found in Table 1.

Individual patient data were provided by trial authors for the remaining five trials which recruited a total of 669 participants, representing 60% of individuals from all 1119 identified eligible trials (Craig 1994; De Silva 1996; Heller 1995;Ramsay 1992;Turnbull 1985). Data were converted from paper format to computer datasets in two trials (Ramsay 1992; Turnbull 1985), computerised data were provided directly in one trial (Craig 1994) and a combination of both (although mostly computerised) were supplied by the authors of two trials (De Silva 1996; Heller 1995).

Table 1. Outcomes considered and summary of results for trials with no IPD
TrialOutcomes reportedSummary of results
Callaghan 1985

1. Seizure control:

(a) Excellent (seizure-free)

(b) Good (> 50% reduction)

(c) Poor (< 50% re duction)

2. Adverse events

1. PHT (n = 58) SV (n = 64)

(a) 39 (67%)       34 (53%)

(b) 7 (12%)        16 (25%)

(c) 12 (21%)        14 (22%)

2. 6 (10%)         7 (11%)

Czapinski 1997a

1.Proportion achieving 24-month remission at 3 years

2.Proportion excluded after randomisation due to adverse events or no efficacy.

1. PHT: 59% SV: 64% .

2. PHT: 23% SV: 23%

Forsythe 1991

1. Cognitive assessments.

2. Withdrawals from randomised drug. .

1. Significant difference favouring SV test of speed of information processing (P < 0.01).

No significant differences between treatment groups for any other cognitive tests.

2. PHT: 6/20 (30%)  SV:    7/21 (33%)

Rastogi 1991

1.Reduction in frequency of seizures at 24 weeks

(a) Excellent (100% reduction)

(b) Good (75 - 99% reduction)

(c) Fair (50 - 74% reduction)

(d) Poor (< 50% reduction)

2.Adverse events

1. PHT (n = 45) SV (n = 49)

(a) 23 (51%)       24 (49%)

(b) 13 (24%)        17 (35%)

(c) 8 (18%)        5(10%)

(d) 1 (2%)         3 (6%)

2. All reported adverse events were minor.

PHT: gum hyperplasia (18%), nystagmus (13%), gastrointestinal symptoms (4%), drowsiness (4%),  ataxia (2%).

SV: gastrointestinal symptoms (12%), drowsiness (6%),  weight gain (2%)

Shakir 1981

1.Seizures during treatment

2.Adverse events

1. PHT: 5 (33%) SV: 7 (39%)

2. PHT: 1 case of ataxia, 5 cases of acne. SV: 2 cases of gastrointestinal symptoms, 2 cases of hair loss, 4 cases of weight gain.

Thilothammal 1996

1.Recurrence of seizures

2.Adverse events

1. PHT: 14/52 (27%)                 SV: 10/48 (21%).

2. PHT: 33/52 (63%)                 SV: 15/48 (31%)

Data were available for the following participant characteristics (percentage of participants with data available): gender (100%); seizure type (100%); age at randomisation (99%); number of seizures in the six months prior to randomisation (79%); time since first seizure to randomisation (73%). Electroencephalographic (EEG) data had been recorded for all five trials, but only computerised in two trials (Craig 1994; Turnbull 1985;). Similar difficulties were encountered with computerised tomography/magnetic resonance imaging (CT/MRI) data, and neurological examination findings.

One trial (Ramsay 1992) recruited only individuals with generalised onset tonic-clonic seizures, some of whom were experiencing other generalised seizure types such as absence or myoclonus. All generalised seizure types were recorded during follow-up for this trial. The remaining four trials recruited individuals with partial onset seizures (simple/complex partial or secondarily generalised tonic-clonic) and individuals with generalised onset tonic-clonic seizures. For the individuals with generalised onset tonic-clonic seizures recruited into these four trials, other generalised seizure types were not recorded during follow-up. As a result, the majority of the data from the five trials does not address the treatment of generalised seizure types such as absence or myoclonus but applies only to generalised onset tonic-clonic seizures. In our primary analysis, we use only the data for generalised onset tonic-clonic seizures during follow-up as this is the most consistent approach; we also report a sensitivity analysis which includes data on all generalised seizure types from Ramsay 1992.

Excluded studies

We excluded 14 duplicate trials (Berg 1993; Callaghan 1981; Callaghan 1983; Callaghan 1984; Craig 1993; Czapinski 1997b; Czapinski 1997c; Goggin 1984; Goggin 1986; Shakir 1980; Tallis 1994a; Tallis 1994b; Turnbull 1982; Wilder 1983) and we retained the most relevant primary reference for each trial in the review. One trial was not randomised (Zeng 2010) and four (Jannuzzi 2000; Kaminow 2003; Sabers 1995; Schmidt 2007) did not make a randomised comparison between phenytoin (PHT) and valproate (SV) (see Characteristics of excluded studies for detailed reasons for exclusion).

Risk of bias in included studies

For further details see Characteristics of included studies, Figure 2 and Figure 3.

Figure 2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

(1) Trials for which individual patient data were provided

Three trials reported adequate methods of randomisation and allocation concealment; two trials used permuted blocks to generate a random list and concealed allocation by using sealed opaque envelopes (De Silva 1996; Heller 1995). One trial used a computer minimisation programme and a pharmacy-controlled allocation (Craig 1994). Two trials did not report sufficient information about methods of randomisation and allocation concealment (Ramsay 1992; Turnbull 1985).

(2) Trials for which no individual patient data were available

Two trials reported adequate methods of randomisation: telephone randomisation (Shakir 1981) and a computer-generated list of randomised numbers (Thilothammal 1996). Two trials reported no information on methods of randomisation (Czapinski 1997a; Rastogi 1991), one trial reported unclear information on randomisation (Callaghan 1985) and one trial reported an inadequate method of randomisation, i.e. quota allocation (Forsythe 1991). Five of the six trials reported no information on allocation concealment (Czapinski 1997a; Forsythe 1991; Rastogi 1991; Shakir 1981; Thilothammal 1996) and one trial reported an inadequate method of allocation concealment based on 'drug of first preference' (Callaghan 1985).

Blinding

(1) Trials for which individual patient data were provided

One trial was single-blinded (outcome assessor for cognitive testing) (Craig 1994), three trials were unblinded for "practical and ethical reasons" (De Silva 1996; Heller 1995; Ramsay 1992) and one trial provided no information on blinding (Turnbull 1985).

(2) Trials for which no individual patient data were available

One trial was double-blinded (Thilothammal 1996), one trial was single-blinded (outcome assessor for cognitive testing) (Forsythe 1991) and no information was provided on blinding in the other trials (Callaghan 1985; Czapinski 1997a; Rastogi 1991; Shakir 1981).

Incomplete outcome data

(1) Trials for which individual patient data were provided

In theory, a review using individual patient data should overcome issues of attrition bias as unpublished data can be provided, unpublished outcomes calculated and all randomised participants can be analysed by an intention-to-treat approach. All five trials (Craig 1994; De Silva 1996; Heller 1995; Ramsay 1992; Turnbull 1985) reported attrition rates and provided individual patient data for all randomised individuals.

(2) Trials for which no individual patient data were available

Four trials reported attrition rates and analysed all randomised participants using an intention-to-treat approach (Callaghan 1985; Forsythe 1991; Shakir 1981; Thilothammal 1996). Two trials did not provide sufficient information to assess attrition bias (Czapinski 1997a; Rastogi 1991).

Selective reporting

The authors of Craig 1994 provided a protocol; the outcomes specified in the protocol were consistent with the outcomes reported in the publication, and we therefore judged the risk of selective reporting bias to be low. Protocols were not available for any of the other ten included trials so we made a judgement of the risk of bias based on the information included in the publications (see Characteristics of included studies for more information).

(1) Trials for which individual patient data were provided

In theory, a review using individual participant data should overcome issues of reporting biases as unpublished data can be provided and unpublished outcomes calculated. Sufficient individual patient data were provided to calculate the four outcomes ('Time to withdrawal of allocated treatment', 'Time to six-month remission', 'Time to 12-month remission' and 'Time to first seizure') for four of the five trials (De Silva 1996; Heller 1995; Ramsay 1992; Turnbull 1985). Withdrawal information was not provided for one trial (Craig 1994), so that 'Time to withdrawal of allocated treatment' could not be calculated but sufficient information was provided to calculate the other three outcomes.

(2) Trials for which no individual patient data were available

Seizure outcomes and adverse events were well reported in four trials (Callaghan 1985; Rastogi 1991; Shakir 1981; Thilothammal 1996). One trial reported cognitive outcomes and adverse events but no seizure outcomes (Forsythe 1991); however as no protocol was available for this trial we do not know whether seizure outcomes were planned a priori. One trial was in abstract form only and did not provide sufficient information to assess selective reporting bias (Czapinski 1997a).

Other potential sources of bias

We detected no other potential sources of bias in any of the 11 trials included in the review

Effects of interventions

See: Summary of findings for the main comparison

A summary of the outcomes reported in trials for which no individual patient data (IPD) were available are reported in Table 1. Details regarding the number of individuals (with IPD) contributing to each analysis are given in Table 2. All results are summarised in Table 3 and Meta View. Survival curve plots are shown in Figure 4; Figure 5; Figure 6; Figure 7; Figure 8; Figure 9; Figure 10; Figure 11. All survival curve plots were produced in Stata software version 11.2 (Stata 2009) using data from all trials providing IPD combined. We would have liked to stratify by trial in survival curve plots but we do not know of any software which allows for this; we hope that such software may have been developed for future updates of this review.

Figure 4.

Time to withdrawal of allocated treatment

One participant randomised to phenytoin (PHT) and nine participants randomised to valproate (SV) had time to withdrawal of zero days and are therefore not included in "Number at Risk"

Figure 5.

Time to withdrawal of allocated treatment - stratified by epilepsy type

One participant with generalised epilepsy randomised to phenytoin (PHT) and nine participants with generalised epilepsy randomised to valproate (SV) had time to withdrawal of zero days and are therefore not included in "Number at Risk"

Figure 6.

Time to 12-month remission

Figure 7.

Time to 12-month remission - stratified by epilepsy type

Figure 8.

Time to six-month remission

Figure 9.

Time to six-month remission - stratified by epilepsy type

Figure 10.

Time to first seizure

Figure 11.

Time to first seizure - stratified by epilepsy type

Table 2. Number of individuals contributing to each analysis
  1. 1 Withdrawal information not provided for Craig 1994 so cannot contribute to 'Time to withdrawal of allocated treatment'

    2 Follow-up for Ramsay 1992 is less than 12 months so cannot contribute to 'Time to 12-month remission'

    3 Data extracted from Forsythe 1991 and Shakir 1981 publications to calculate time to withdrawal of allocated treatment. Insufficient published data to calculate other outcomes

    PHT = Phenytoin; SV = Valproate

TrialNumber randomised

Time to withdrawal of

allocated treatment

Time to 12 month remissionTime to 6 month remissionTime to first seizure
PHTSVTotalPHTSVTotalPHTSVTotalPHTSVTotalPHTSVTotal
Craig 1994 18185166000717614771761477176147
De Silva 199654491035347119544910354491035449103
Forsythe 1991 3202141202141000000000
Heller 199563611246158100636112463611246361124
Ramsay 1992 25086136508613600048771254877125
Turnbull 198570701407070140707014070701407070140
Shakir 1981 3151833151833000000000
Total353390743269300569258256514306333639306333639
Table 3. Results of Analysis (heterogeneity, overall effect and interaction)
  1. CI: confidence interval
    df: degrees of freedom of Chi² distribution
    P values < 0.05 are classified as statistically significant
    HR: Hazard ratio

  Statistic

Time to withdrawal of

allocated treatment

Time to 12 month

remission

Time to six month

remission

Time to first seizure
Test for heterogeneityChi²(df = 5) 5.95(df = 3) 0.19(df = 4) 1.66(df = 4) 4.23
P value0.310.980.800.38
16%0%0%5%
Overall effectHR (95% CI)1.02 (0.73 to 1.49)0.97 (0.77 to 1.22)0.92 (0.76 to 1.12)0.96 (0.78 to 1.18)
P value0.920.810.42 0.70

Test for interaction between

treatment and epilepsy type

Chi²(df = 1) 0.31(df = 1) 0.39(df = 1) 0.13(df = 1) 1.06
P value0.580.530.720.3
0%0%0%5.6%

Overall effect adjusted for

epilepsy type

HR (95% CI)1.09 (0.76 to 1.55)0.98 (0.78 to 1.23)0.95 (0.78 to 1.15)0.93 (0.75 to 1.14)
P value0.190.870.600.47

All hazard ratios (HRs) presented below are calculated by generic inverse variance fixed-effect meta-analysis unless otherwise stated.

(1)Time to withdrawal of allocated treatment

For this outcome, a HR greater than one indicates a clinical advantage for valproate.

Unadjusted analysis

Time to withdrawal of allocated treatment and reason for withdrawal were available for 495 individuals from four trials (De Silva 1996; Heller 1995; Ramsay 1992; Turnbull 1985); 74% of individuals from five trials providing IPD (44% of all 1119 eligible individuals). Withdrawal data were not available for the fifth trial (Craig 1994). Sufficient IPD were available in the trial publications for a further 74 individuals from two trials (Forsythe 1991; Shakir 1981). Therefore a total of 569 individuals (51% of 1119 eligible individuals) from six trials could contribute to the analysis of this outcome; 146 (26%) prematurely withdrew from treatment, 72/269 participants (27%) randomised to phenytoin and 74/300 participants (25%) randomised to valproate. See Table 4 for reasons for premature termination of the study by treatment and how we classified these withdrawals in analysis.

Table 4. Reasons for premature discontinuation (withdrawal of allocated treatment)
  1. n = number of individuals contributing to the outcome 'Time to treatment withdrawal'
    1 IPD for 'Time to treatment withdrawal' was not provided for Craig 1994
    2 Three participants for Heller 1995 (all SV) and three for De Silva 1996 (one PHT and two SV) have missing reasons for treatment withdrawal
    3 Four participants from Heller 1995 had missing withdrawal times and did not contribute to analysis but reasons for withdrawal are given
    4 Death due to reasons not related to the study drug
    5 Other reasons from Ramsay 1992 – two participants withdrew due to pregnancy and one for personal reasons

Reason for early terminationClassificationDe Silva 19962Heller 19952,3Ramsey 1992Turnbull 1985Total1

PHT

n = 53

SV

n = 47

PHT

n = 63

SV

n = 58

PHT

n = 50

SV

n = 86

PHT

n = 70

SV

n = 70

PHT

n = 236

SV

n = 261

Adverse events/intoxicationEvent2214571472220
Poor seizure control/lack of efficacyEvent10118921022023
Both adverse events and lack of efficacyEvent54260021911
Non-complianceEvent0000172239
Participant went into remissionCensored2416141300003829
Lost to follow-upCensored0000410771117
Death4Censored0000003333
Other5Censored0000210021
Completed the study/did not withdrawCensored1214382636604248128148

The overall pooled HR (for 569 participants)  is 1.02 (95% confidence interval (CI) 0.73 to 1.42, P = 0.92) indicating no clear advantage for either drug . There is no evidence of statistical heterogeneity between trials (Chi² = 5.95, df = 5, P = 0.31, I² = 16%) (see Analysis 1.1).  

Table 4 shows that 29/128 (23%) participants on phenytoin and 31/148 (21%) participants on valproate withdrew from the study due to adverse events (or a combination of lack of efficacy and adverse events). See Table 5 for details of all adverse event data provided in the studies included in this review. In summary, the adverse events reported by two or more studies in this review are:

Table 5. Adverse event data (narrative report)
  1. 1  Adverse event data as reported narratively in the publications. Adverse event data were not requested in original IPD requests but will be for all future IPD requests. For numbers of withdrawals due to adverse events in studies for which IPD were provided (De Silva 1996; Heller 1995; Ramsay 1992; Turnbull 1985) see Table 4

    2 Participants may report more than one adverse event.

    3 The published paper Craig 1994 reports on a subset of 38 participants, so the adverse event data summary applies only to this subset. IPD were provided for 166 participants (no additional adverse event data provided)

    4 Czapinski 1997 is an abstract only so very little information is reported

    5 Participants may have withdrawn due to adverse event alone or a combination of adverse events and poor efficacy (seizures)

    6 Most commonly reported adverse events only, no indication of overall frequency of all adverse events

TrialAdverse Event Data1Summary of reported results
Phenytoin (PHT)SV (Sodium Valproate)
Callaghan 1985All adverse events developed (by drug) and adverse events leading to discontinuation of treatment

PHT (n = 58): gum hypertrophy (n = 2), rash (n = 2), ataxia (n = 2)

 

SV (n = 64): weight gain (n = 4 – all discontinued treatment), drowsiness (n = 2), aggressive behaviour (n = 1 – discontinued treatment)
Craig 1994

Adverse event frequency (spontaneous reports)2

Discontinuations due to adverse events3

PHT (n = 25): unsteadiness (n = 9), sleepiness (n = 7), drowsiness (n = 2), impaired concentration (n = 2), confusion (n = 1), constipation (n = 1), diarrhoea (n = 1), dysarthria (n = 1), lethargy (n = 1), nystagmus (n = 1), rash (n = 1), tired legs (n = 1)

PHT discontinuations (n = 6): rash (n =1 ), diarrhoea (n = 1), confusion (n = 1), unsteadiness (n = 1), constipation (n = 1), sleepiness (n = 1)

SV (n = 17): unsteadiness (n = 2), sleepiness (n = 3), tremor (n = 5), oedema (n = 3), alopecia (n = 2), depression (n = 2), weight gain (n = 2)

SV discontinuations (n = 2): weight gain and depression (n = 1), unsteadiness (n =1)

Czapinski 1997a“Exclusions” due to adverse events or no efficacy4Proportion “excluded”: PHT : 33.3%Proportion “excluded”: SV: 23.3%
De Silva 1996“Unacceptable” adverse events leading to drug withdrawal5PHT (n = 54): drowsiness (n = 2), skin rash (n = 1) blood dyscrasia (n = 1), hirsutism (n = 1)SV (n = 49): behavioural (n = 1), tremor (n = 1)
Forsythe 1991

No adverse event data reported.

(Withdrawal data only reported)

1 participant (PHT) withdrew from the study due to depression and anorexiaNo adverse event data (or withdrawals due to adverse events) reported
Heller 1995“Unacceptable” adverse events leading to drug withdrawal5

PHT (n = 63): myalgia (n = 1), irritability (n = 1)

 

SV (n = 61): dizziness (n = 2) abnormal liver function test (n = 1)
Ramsay 1992Most common adverse events (by treatment group)6

PHT (n = 50): dyspepsia (n = 1), nausea (n = 2), dizziness (n = 2), somnolence (n = 5), tremor (n = 2), rash (n = 4)

 

SV (n = 86): dyspepsia (n = 7), nausea (n = 10), dizziness (n = 5), somnolence (n = 8), tremor (n = 5), rash (n = 3)
Rastogi 1991Commonest adverse events (reported as percentages by treatment group)6PHT (n = 45): gum hyperplasia (17.7%), nystagmus (13.33%), ataxia (2.2%), gastrointestinal disturbances (4.44%), drowsiness (4.44%)SV (n = 49): gastrointestinal disturbances (12%), drowsiness (6.12%), weight gain (2.04%)
Shakir 1981Adverse events (narrative description)2PHT (n = 15): 1 case of ataxia, 5 cases of acne.SV (n = 18): 2 cases of gastrointestinal symptoms, 2 cases of hair loss, 4 cases of weight gain
Thilothammal 1996Assessment of adverse events2

PHT (n = 52): 33 participants reported at least one side effect.

Reported frequencies: gingival hypertrophy (n = 30), ataxia (n = 13), sedation (n = 12), nausea and vomiting (n = 1)

Other reported adverse events (no frequencies): nystagmus, confusion

SV (n = 48): 15 participants reported at least one side effect.

Reported frequencies: hyperactivity (n = 6), impaired school performance (n = 4), severe skin allergy (n = 1)

Turnbull 1985Withdrawals due to dose-related and idiosyncratic adverse events

PHT (n = 70) 11 withdrawals due to dose-related adverse events (nystagmus, ataxia, tremor, diplopia and mental change)

5 withdrawals due to idiosyncratic adverse events (skin eruption, erythroderma and jaundice)

SV (n = 70) 9 withdrawals due to dose-related adverse events (tremor, irritability, restlessness and alopecia). No withdrawals due to idiosyncratic adverse events

For valproate

For phenytoin

It is difficult to summarise the 'most common' adverse events overall across the 11 studies due to the differences in methods of reporting adverse event data across the studies (see Table 5 for more information). Adverse event data for individuals were not included in the original IPD requests for earlier versions of this review but will be pledged in all future IPD requests.

Adjusted analysis

Withdrawal data for 41 participants extracted from Forsythe 1991 did not distinguish between seizure type (partial onset or generalised onset) and therefore could not be included in the meta-analysis stratified by seizure type.

The overall pooled HR (adjusted by seizure type for 528 participants) is 1.09 (95% CI 0.76 to 1.55, P = 0.64) indicating a slight advantage for valproate which is not statistically significant (see Analysis 1.2). This result is similar to the unadjusted pooled HR (Analysis 1.1) and conclusions remain unchanged following the exclusion of 41 individuals (Forsythe 1991) in the stratified analysis. For individuals with generalised onset seizures (341), the pooled HR was 0.98 (95% CI 0.59 to 1.64, P = 0.94) indicating no clear advantage for either drug. For individuals with partial onset seizures (187), the pooled HR was 1.20 (95% CI 0.74 to 1.95, P = 0.47) suggesting an advantage for valproate which is not statistically significant. There was no evidence of an interaction between epilepsy type (partial onset versus generalised onset) and treatment effect (Chi² = 0.31, df = 1, P = 0.58, I² = 0%). (See Analysis 1.2)

An important amount of heterogeneity was present between trials within the generalised onset seizure subgroup (Chi² = 8.44, df = 4, P = 0.08, I² = 53%). On visual inspection of the forest plot (see Analysis 1.2) one trial (Heller 1995) appears to be the source of this variability as this trial shows a large statistically significant treatment effect in favour of phenytoin while the other four trials show a general non-significant trend in favour of valproate (Ramsay 1992; Shakir 1981; Turnbull 1985).

In Ramsay 1992, there is an indication that the proportional hazards assumption may be violated (see Data synthesis); the P value of time-varying covariate is 0.054; however visual inspection of the survival plot (Figure 12) shows no indication of survival curves crossing which would imply non-proportional hazards. As a sensitivity analysis, a piecewise Cox regression model is fitted to investigate any change in treatment effect over time assuming proportional hazards within each interval. The follow-up period of Ramsay 1992 is split into three intervals based on the number of events and number of individuals at risk in each interval; time to withdrawal of allocated treatment occurring 0 to 50 days, 51 to 100 days or after 100 days. Separate hazard ratios can be estimated for each interval:

Figure 12.

Time to withdrawal of allocated treatment - Ramsay 1992

(I) For interval 0 to 50 days (13 events from 128 participants at risk) the HR is 2.79 (95% CI 0.91 to 8.54, P = 0.07) suggesting a large advantage for valproate which is not statistically significant

(ii) For interval 51 to 100 days (6 events from 110 participants at risk) the HR is 0.95 (95% CI 0.41 to 2.23, P = 0.91) suggesting no clear advantage for either drug.

(iii) For interval after 100 days (4 events from 102 participants at risk) the HR is 0.83 (95% CI 0.39 to 1.77, P = 0.63) suggesting an advantage for phenytoin which is not statistically significant.

These results suggest some indication of a change in treatment effect over time with phenytoin more likely to be withdrawn early and valproate more likely to be withdrawn later; however, the confidence intervals of the estimates are wide due to the small numbers of events within each interval so there is insufficient information to support the hypothesis of a change in treatment effect over time for Ramsay 1992. However, this study has a shorter length of follow-up than the other studies included for this outcome (median time to treatment withdrawal for Ramsay 1992 is 180.5 days and is 815 days, 952 days, 851 days and 912 days for De Silva 1996, Heller 1995, Shakir 1981 and Turnbull 1985 respectively). The length of follow-up may therefore be too short to examine the hypothesis of a change in treatment effect over time.

(2) Time to achieve 12-month remission

For this outcome, a HR greater than one indicates a clinical advantage for phenytoin.

Data for 514 individuals (77% of those providing IPD) from four trials (Craig 1994; De Silva 1996; Heller 1995; Turnbull 1985) were available for the analysis of this outcome. Individuals were only followed up for six months in the fifth trial (Ramsay 1992), which could not contribute data to this outcome.

The overall pooled HR (for 514 participants) is 0.97 (95% CI 0.77 to 1.22, P = 0.81) indicating no clear advantage to either drug. There is no evidence of statistical heterogeneity between trials (Chi² = 0.19, df = 3, P = 0.98, I² = 0%) (see Analysis 1.3). For individuals with generalised seizures (270), the pooled HR was 1.04 (95% CI 0.77 to 1.40, P = 0.79 ) indicating no clear advantage for either drug. For individuals with partial onset seizures (244), the pooled HR was 0.90 (95% CI 0.63 to 1.29, P = 0.56) indicating an advantage for valproate which is not statistically significant. Overall, the pooled HR (adjusted for seizure type for 514 participants) was 0.98 (95% CI 0.78 to 1.23, P = 0.87) suggesting no clear clinical advantage for either drug. There was no evidence of an interaction between epilepsy type (partial onset versus generalised onset) and treatment (Chi² = 0.39, df = 1, P = 0.53) (see Analysis 1.4).

The proportional hazards assumption of the Cox model was satisfied for all trials.

(3) Time to achieve six-month remission

For this outcome, a HR greater than one indicates a clinical advantage for phenytoin.

Data for 639 individuals (96% of those providing IPD) from five trials were available for the analysis of this outcome.

The overall pooled HR (for 639 participants) is 0.92 (95% CI 0.76 to 1.12, P = 0.42) indicating an advantage of valproate which is not statistically significant. There is no evidence of statistical heterogeneity between trials (Chi² = 1.66, df = 4, P = 0.80, I² = 0%) (see Analysis 1.5). For individuals with generalised seizures (395), the pooled HR was 0.92 (95% CI 0.72 to 1.18, P = 0.53), suggesting an advantage for valproate which is not statistically significant. For individuals with partial onset seizures (244), the pooled HR was 0.99 (95% CI 0.73 to 1.35, P = 0.96) indicating no clear advantage for either drug. Overall, the pooled HR (adjusted for seizure type for 639 participants) was 0.95 (95% CI 0.78 to 1.15, P = 0.60) suggesting no clear advantage for either drug. There was no evidence of an interaction between epilepsy type (partial onset versus generalised onset) and treatment (Chi² = 0.13, df = 1, P = 0.72) (see Analysis 1.6).

The proportional hazards assumption of the Cox model was satisfied for all trials.

A sensitivity analysis including generalised seizures of all types during follow-up (only recorded in Ramsay 1992) produced the following results: for individuals with generalised seizures (395), the pooled HR was 0.84 (95% CI 0.62 to 1.14, P = 0.26), suggesting an advantage for valproate which is not statistically significant. For individuals with partial onset seizures (244), the pooled HR was unchanged, 0.99 (95% CI 0.73 to 1.35, P = 0.96) indicating no clear advantage for either drug. Overall, the pooled HR (adjusted for seizure type) was 0.91 (95% CI 0.73 to 1.13, P = 0.40), suggesting an advantage for valproate which is not statistically significant

By including information on other generalised seizure types in the trial by Ramsay 1992, a slightly greater advantage for valproate emerges. As the overall results from both analyses are similar and overall conclusions are unchanged, we will focus on the original analysis which includes only generalised tonic-clonic seizures during follow-up in all trials.

(4) Time to first seizure post randomisation

For this outcome, a HR greater than one indicates a clinical advantage for valproate.

Data for 639 individuals (96% of those providing IPD) from five trials were available for the analysis of this outcome.

The overall pooled HR (for 639 participants) is 0.96 (95% CI 0.78 to 1.18, P = 0.70) indicating no clear advantage for either drug. There is no evidence of statistical heterogeneity between trials (Chi² = 4.23, df = 4, P = 0.38, I² = 5%) (see Analysis 1.7).

For individuals with generalised seizures (395), the pooled HR was 1.03 (95% CI 0.77 to 1.39, P = 0.82) indicating no clear advantage for either drug. For individuals with partial onset seizures (244), the pooled HR was 0.83 (95% CI 0.62 to 1.11, P = 0.22) suggesting an advantage for phenytoin which is not statistically significant. Overall, the pooled HR (adjusted for seizure type for 639 participants) was 0.93 (95% CI 0.75 to 1.14, P = 0.45) suggesting an advantage for phenytoin which does not reach statistical significance. There was no evidence of an interaction between epilepsy type (partial onset versus generalised onset) and treatment effect (Chi² = 1.06, df = 1, P = 0.03) (see Analysis 1.8).

The proportional hazards assumption of the Cox model was satisfied for all trial-specific estimates of the log(hazard ratio).

A sensitivity analysis including generalised seizures of all types during follow-up (only recorded in Ramsay 1992) produced the following results: for individuals with generalised seizures, the pooled HR was 1.05 (95% CI 0.79 to 1.40, P = 0.74), indicating no clear advantage for either drug. For individuals with partial onset seizures, the pooled HR was unchanged, 0.83 (95% CI 0.62 to 1.11, P = 0.22) , suggesting an advantage for phenytoin which is not statistically significant. Overall, the pooled HR (adjusted for seizure type) was 0.93 (95% CI 0.76 to 1.15, P = 0.52), suggesting an advantage for phenytoin which is not statistically significant.

As the overall results from both analyses are similar and overall conclusions are unchanged, we will focus on the original analysis which includes only generalised tonic-clonic seizures during follow-up in all trials.

(5) Misclassification of seizure type
We did not find evidence of an interaction between treatment and seizure type in any analysis. This result is surprising, given the strong clinical impression that valproate is more effective in generalised onset seizures while phenytoin is more effective in partial onset seizures. The impression that valproate is better for generalised seizures may derive from its effects on generalised seizures other than tonic-clonic; however, we were unable to investigate these seizure types in this review. Misclassification of seizure type (whereby some individuals with generalised seizures have been mistakenly classed as having partial onset seizures, and vice versa) is a well-recognised problem in epilepsy, and it may be that an interaction between treatment and seizure type has been masked because of this. Given clinical evidence that individuals with generalised onset seizures are unlikely to have an 'age at onset' greater than between 25 and 30 years (Malafosse 1994), we examined the distribution of age at onset for individuals with generalised seizures. This revealed that a substantial number of individuals classified as having generalised seizures had an age at onset over 30 years:

Therefore a total of up to 190 out of 384 individuals (49%) classified as having generalised onset seizures may have had their seizure type mis classified. Such a misclassification could bias our results against finding an interaction between treatment and seizure types (partial onset versus generalised onset) and could explain why we have not found strong evidence to support the clinical impression that such an interaction exists. We decided to investigate this further.

We undertook the following two analyses to investigate misclassification:
(i) We reclassified all individuals with generalised seizures and age at onset greater than 30 into an 'uncertain seizure type' group;
(ii) We reclassified individuals with generalised seizures and age at onset greater than 30 as having partial onset seizures.

The results for each outcome are summarised in Table 6.

Table 6. Sensitivity analysis - Epilepsy type misclassification, fixed-effect analysis
  1. P: Partial epilepsy, G: Generalised epilepsy, O: Overall (all participants). Results are presented as pooled HR (95% CI) with fixed-effect
    Chi² : Chi² statistic. df: degrees of freedom of Chi² distribution
    P: P value. (< 0.05 are classified as statistically significant)
    100 participants reclassified to partial epilepsy or uncertain epilepsy type for outcome 'Time to treatment withdrawal'
    145 participants reclassified to partial epilepsy or uncertain epilepsy type for outcome 'Time to 12-month remission'
    171 participants reclassified to partial epilepsy or uncertain epilepsy type for outcome 'Time to 6-month remission' or 'Time to first seizure'

    See Analysis 1.2; Analysis 1.4; Analysis 1.6; and Analysis 1.8 for original analyses of 'Time to treatment withdrawal', 'Time to 12-month remission', 'Time to 6-month remission' and 'Time to first seizure' respectively.

    See Analysis 1.9 and Analysis 1.10 for forest plots of 'Time to first seizure' sensitivity analyses for generalised and age at onset > 30 reclassified as uncertain epilepsy type and partial epilepsy respectively. Forest plots are not presented for 'Time to treatment withdrawal', 'Time to 6-month remission', 'Time to 12-month remission' sensitivity analyses as results were similar for partial onset and generalised onset subgroups and conclusions unchanged.

 

Time to withdrawal of

allocated treatment

Time to 12-month remission Time to 6-month remission Time to first seizure
(i) Original analysis

P: 1.20 (0.76 to 1.95)

G: 0.98 (0.59 to 1.64)

O: 1.09 (0.76 to 1.55)

P: 0.90 (0.63 to 1.29)

G: 1.04 (0.77 to 1.40)

O: 0.98 (0.78 to 1.23)

P: 0.99 (0.73 to 1.35)

G: 0.92 (0.72 to 1.18)

O: 0.95 (0.78 to 1.15)

P: 0.83 (0.62 to 1.11)

G: 1.03 (0.77 to 1.39)

O: 0.93 (0.75 to 1.14).

(i) Test for interaction

Chi² = 0.31, df = 1,

P = 0.58, I² = 0%

Chi² = 0.39, df = 1,

P =0.53, I² = 0%

Chi² = 0.13, df = 1,

P =0.72, I² = 0%

Chi² = 1.06, df = 1,

P = 0.30, I² = 5.6%

     

(ii) Generalised onset and age at onset > 30 classified as

uncertain seizure type

P: 1.20 (0.76 to 1.95)

G: 1.33 (0.74 to 2.38)

U: 0.47 (0.12 to 1.85)

O: 1.17 (0.82 to 1.67)

P: 0.90 (0.63 to 1.29)

G: 0.93 (0.63 to 1.39)

U: 1.36 (0.85 to 2.17)

O: 1.01 (0.80 to 1.27)

P: 0.99 (0.73 to 1.35)

G: 0.88 (0.62 to 1.25)

U: 1.11 (0.76 to 1.61)

O: 0.99 (0.81 to 1.20)

P: 0.83 (0.62 to 1.11)

G: 1.34 (0.91 to 1.97)

U: 0.74 (0.47 to 1.17)

O: 0.93 (0.76 to 1.15)

(ii) Test for interaction

Chi² = 1.88, df = 2,

P =0.39, I² = 0%

Chi² = 2.07, df = 2,

P = 0.36, I² = 3.3%

Chi² = 0.78, df = 2,

P = 0.68, I² = 0%

Chi² = 4.87, df = 2,

P = 0.09, I²= 58.9%

     
(iii) Generalised onset and age at onset > 30 reclassified as partial onset

P: 0.98 (0.63 to 1.53)

G: 1.33 (0.74 to 2.38)

O: 1.10 (0.77 to 1.56)

P: 1.01 (0.76 to 1.34)

G: 0.93 (0.63 to 1.39)

O: 0.98 (0.78 to 1.24)

P: 1.00 (0.79 to 1.27)

G: 0.88 (0.62 to 1.25)

O: 0.97 (0.80 to 1.18)

P: 0.81 (0.64 to 1.04)

G: 1.34 (0.91 to 1.97)

O 0.94 (0.76 to 1.15)

(iii) Test for interaction

Chi² = 0.67, df = 1,

P = 0.41, I² = 0%

Chi² = 0.10, df = 1,

P = 0.75, I² = 0%

Chi² = 0.36, df = 1

P = 0.55, I² = 0%

Chi² = 4.55, df = 1

P = 0.03, I² = 78%

  • For 'Time to withdrawal of allocated treatment', reclassifying individuals does not provide stronger evidence of an interaction. However, results for the 'uncertain seizure type' subgroup (pooled HR of 0.47 (95% CI 0.12 to 1.85) indicating a large but non-significant advantage for phenytoin) are substantially different in the direction of effect from estimates for the 'partial onset seizures' subgroup (pooled HR of 1.20 (95% CI 0.74 to 1.95)) and generalised onset seizure groups (pooled HR of 1.33 (95% CI 0.74 to 2.38)) both indicating a non-significant advantage for valproate.

  • Similarly, for 'Time to 12-month remission', reclassifying individuals does not provide stronger evidence of an interaction and again results for the 'uncertain seizure type' subgroup (pooled HR of 1.36 (95% CI 0.85 to 2.17) indicating a non-significant advantage for phenytoin) are substantially different in the direction of effect from estimates for the 'partial onset seizures' subgroup (pooled HR of 0.90 (95% CI 0.63 to 1.29)) and generalised onset seizure groups (pooled HR of 0.93 (95% CI 0.63 to 1.39)) both indicating a non-significant advantage for valproate.

  • The results for 'Time to six-month remission' are very similar regardless of whether individuals have been reclassified or not.

  • For 'Time to first seizure', reclassifying individuals results in a more obvious interaction between treatment and seizure type:

    • For generalised seizures and age of onset more than 30 reclassified as 'uncertain seizure type', the result of the test of interaction between treatment and seizure type is Chi² = 4.87, df = 2, P = 0.09 (see Analysis 1.9)

    • For generalised seizures and age of onset more than 30 reclassified as 'partial onset seizures', the result of the test of interaction between treatment and seizure type is Chi² = 4.55, df = 1, P = 0.03 (see Analysis 1.10)

The direction of effect for the 'uncertain seizure type' subgroup (pooled HR of 0.74 (95% CI 0.47 to 1.17)) is similar to that of the 'partial onset' subgroup (pooled HR of 0.83 (95% CI 0.62 to 1.11)) both indicating a non-significant advantage for phenytoin. Valproate now appears even more effective in generalised onset seizures (pooled HR of 1.34 (95% CI 0.91 to 1.97)) when compared to the original analysis (see Analysis 1.9 and Analysis 1.10).

Reclassifying seizure type on the basis of age of onset provides evidence for an interaction between treatment and seizure type for the outcome 'Time to first seizure' only. A review comparing carbamazepine and valproate monotherapy for epilepsy (Marson 2000) undertook a similar investigation of seizure type misclassification. For the outcome 'Time to 12-month remission', reclassifying seizure type on the basis of age of onset provided evidence of an interaction between seizure type and treatment for that outcome. In that review, however, a significant interaction between age at onset and treatment was also found for 'Time to first seizure'. There were therefore two potential explanations for the interaction found when individuals were reclassified according to age of onset. Firstly, misclassification had masked the interaction between treatment and seizure type in the primary analyses and reclassifying individuals according to age of onset has reduced the bias caused by misclassification. Alternatively, age at onset was acting as an independent predictor of outcome, and the misclassification analysis using age of onset forced the results to reflect this.

To investigate the hypothesis of age at onset acting as an independent predictor of outcome in this review, we performed the following analysis for the outcome 'Time to first seizure':

  1. A Cox Proportional Hazards regression model (stratified by trial) fitted with a single covariate as a treatment indicator (phenytoin = 1, valproate = 0) was fitted.

  2. A Cox Proportional Hazards regression model (stratified by trial) fitted with two covariates, treatment and age (measured as a continuous variable) was fitted.

  3. A Cox Proportional Hazards regression model (stratified by trial) fitted with three covariates, treatment, age and an age*treatment interaction term, was fitted.

The difference in -2(log likelihood) of models 1, 2 and 3 were compared; differences were compared to a Chi² distribution with one or two degrees of freedom. The -2(log likelihood) values of the three models were 3192.808, 3192.410 and 3189.048 respectively. We therefore found no evidence that age is an independent predictor of outcome (P = 0.528 comparing goodness of fit of models 1 and 2). However, there is some indication of an interaction between age at onset and treatment (P = 0.067 comparing goodness of fit of models 2 and 3) as in the carbamazepine versus valproate review (Marson 2000).

In model 3, the effect sizes for the three covariates are as follows:

  • treatment HR 1.25 (95% CI 0.080 to 1.77, P = 0.213) indicating an advantage for valproate which is not statistically significant.

  • age HR 1.002 (95% CI 0.993 to 1.012, P = 0.647) indicating no significant effect of age on outcome.

  • age*treatment interaction HR 0.993 (0.985 to 1.001, P = 0.067) indicating an interaction between treatment and age which is borderline statistically significant. This effect indicates that with increasing age of onset, treatment effect (advantage for valproate) seems to decrease by approximately 0.7% with each increasing year of age of onset. In other words, as in Marson 2000, there is an indication that younger individuals may fare better on valproate than older individuals who fare better on phenytoin for this outcome.

Discussion

Summary of main results

The results of this review do not demonstrate a statistically significant effect in favour of either valproate or phenytoin for the primary global outcome 'Time to withdrawal of allocated treatment'. This outcome is influenced by both the relative efficacy of the two drugs, and differences in tolerability and safety. As a difference in efficacy in one direction may be confounded by a difference in tolerability in the other, it may not be surprising that any estimated differences are small. The confidence intervals for this outcome are too wide to confirm equivalence and clinically important differences have not been excluded, particularly when results for generalised and partial onset seizure subgroups are examined. Furthermore, as at least three of trials contributing IPD to this outcome were open-label, clinical preconceptions about the two treatments such as that valproate is more effective in generalised seizures while phenytoin is more effective in partial onset seizures and lack of masking may have influenced the withdrawal rates of the two treatments.

Similarly for the secondary outcomes 'Time to 12-month remission', 'Time to six-month remission', and 'Time to first seizure', although no statistically significant differences were found between phenytoin and valproate, the confidence intervals are too wide to confirm equivalence.

Overall completeness and applicability of evidence

We have gratefully received individual patient data (IPD) for 669 individuals (60% of individuals from all eligible trials) from the authors of five trials (Craig 1994; De Silva 1996; Heller 1995; Ramsay 1992; Turnbull 1985) which included a comparison of phenytoin with valproate for the treatment of epilepsy. However 376 individuals (33%) from four relevant trials (Callaghan 1985; Czapinski 1997a; Rastogi 1991; Thilothammal 1996) could not be included in any analysis as IPD were not available and outcomes of interest were not reported in the published reports. Sufficient data for 74 individuals (7%) were published in two trials (Forsythe 1991;Shakir 1981) to contribute to analysis for the primary outcome 'Time to withdrawal of allocated treatment' but insufficient data were available to include these individuals in the analyses of other outcomes. Having to exclude data for a third of eligible participants due to lack of individual patient data and insufficient reporting in study publications is likely to impact on the applicability of the evidence, however it is difficult to quantify exactly how large this impact could be.

Quality of the evidence

The five trials for which IPD were made available were of generally good quality with all five trials describing adequate methods of randomisation, and Craig 1994, De Silva 1996 and Heller 1995 also describing adequate methods of allocation concealment. However, none of the five trials described a method of blinding of participants and personnel, and only one trial (Craig 1994) stated that cognitive outcome assessors were blinded to treatment allocation, raising the possibility of performance and detection bias. Three trials (De Silva 1996; Heller 1995; Ramsay 1992) were designed as open-label for "practical and ethical reasons"; for example, Ramsay 1992 stated that the side effects of the respective drugs would "quickly unblind" the trial anyway. A further difference between the five trials was the population recruited; two trials recruited adults of all ages (Heller 1995; Turnbull 1985), one recruited children only (De Silva 1996), one recruited adults and children (Ramsay 1992), and one recruited adults over the age of 60 only (Craig 1994).

An important limitation of the current evidence base is that, of the five trials providing full IPD, only one (Ramsay 1992) collected data on generalised seizure types other than generalised tonic-clonic seizures. Hence, the results for seizure outcomes ('Time to remission' and 'Time to first seizure'), apply only to generalised tonic-clonic seizures, despite the fact that individuals may have been experiencing other generalised seizure types. This problem must be addressed in future trials.

Potential biases in the review process

Examining the subgroup analyses for trends shows inconsistent results. For the primary outcome 'Time to treatment withdrawal', estimates indicate a potentially important advantage for valproate for partial onset seizures, with no clear advantage for either drug for generalised tonic-clonic seizures, which goes against current practice and belief. For 'Time to six-month remission', estimates favour valproate for generalised tonic-clonic seizures, with no clear trend for partial onset seizures. For 'Time to 12-month remission', estimates favour valproate for partial onset seizures, with no clear trend for generalised tonic-clonic seizures, which again contradicts current practice and belief. For 'Time to first seizure', there is a trend in favour of phenytoin for partial onset seizures, with no clear trend for generalised tonic-clonic seizures.

Despite strong prior clinical impressions that valproate is more effective in generalised seizures and that phenytoin is more effective in partial onset seizures, we have failed to detect a significant interaction between treatment and seizure type for any outcome to support current practice. It must however be understood that the confidence intervals around the estimates are wide, and that these results do not exclude the possibility of important differences existing.

Why have we failed to find an interaction between drug and seizure type? It may well be that an interaction does not exist. Alternatively, it may be that an interaction does exist but we have failed to detect it. We suggest the following reasons why this might have occurred.

(1) Our meta-analysis may not have the statistical power needed to detect an interaction.

(2) Generalised tonic-clonic seizures were the only generalised seizure type contributing to the main analyses. It may be that there is no difference between phenytoin and valproate for control of this seizure type, but important differences could exist for absence and myoclonus seizure types. However, were this the case we might have expected to see a treatment-seizure type interaction for the outcome 'Time to treatment withdrawal' if treatment were being withdrawn or a further drug added to combat other seizure types.

(3) Due to the strong clinical impression that valproate is the treatment of choice for individuals with myoclonus and absence seizures, physicians may have been reluctant to randomise individuals with epilepsy syndromes particularly responsive to valproate into these trials (e.g. juvenile myoclonic epilepsy). This seems unlikely given that recruitment into those trials of individuals with generalised tonic-clonic seizures took place some time before such beliefs became widely held in the UK.

(4) The results of the original trials and hence this meta-analysis may have been confounded by classification bias, i.e. individuals with generalised seizures may have been mis classified as having partial onset seizures and vice versa. There is good evidence from a similar review comparing carbamazepine and valproate that misclassification is indeed an important issue in epilepsy trials (Marson 2000). Within our review, the most striking indication that misclassification may be a problem is the classification of subjects in Craig 1994. In this trial, 95 out of 166 (56%) of the recruited individuals were classified as having a generalised epilepsy, which seems unlikely given that the individuals were newly diagnosed and over the age of 60 (Malafosse 1994). It is also interesting to note that Ramsay 1992 is the only trial in this review that attempted to recruit only individuals with generalised tonic-clonic seizures, However, this trial recruited too few individuals to have the power to detect a difference between phenytoin and valproate. In this trial, for a subgroup of individuals with definite electroencephalographic (EEG) changes to support a diagnosis of an idiopathic generalised epilepsy, there appeared to be a greater (but not significant) advantage for valproate, compared to the trial population overall. This could again be interpreted as supporting the potential for misclassification which in turn could confound an interaction between treatment and seizure type. We were unable to test for the effects of EEG changes on the interaction between treatment and seizure type due to EEG data not being collected for all trials, and even where it was available, it was not done in a uniform way. It is likely that these trials were initiated before the publication of the International League Against Epilepsy Classification of Epileptic Syndromes in 1989 (Commission 1989) but they did use the International League Against Epilepsy Classification of Epileptic Seizures that was published in 1981(Commission 1981), which does allow individuals to be classified as those with partial onset or generalised seizures. The age of onset distribution of individuals classified as having generalised seizures indicates misclassification is likely to have occurred in up to 190 out of 384 (49%) of individuals classified as having generalised onset seizures. Our results based on reclassifying the 190 individuals indicate that classification bias is a potentially important confounder of the results of this review, particularly the outcome 'Time to first seizure'. Furthermore, there is evidence for this outcome of an association between age of seizure onset and treatment allocation, suggesting that younger individuals may fare better on valproate while older individuals fare better on phenytoin. For these reasons, it is important that the issue of misclassification is addressed in future trials.

Finally, it should be mentioned that the preparation of valproate used in the included trials may have influenced the results. The trials conducted in the UK (Craig 1994; De Silva 1996; Heller 1995; Turnbull 1985) all used sodium valproate (Epilim). Ramsay 1992, conducted in the USA, used valproic acid (Depakene) which is thought to cause more gastrointestinal side effects than preparations containing either a mixture of sodium valproate and valproic acid, or sodium valproate alone. There is no evidence from randomised controlled trials to support this, but there are some data from observational studies (Brasfield 1999; Cranor 1997; Wilder 1983a). Given that this meta-analysis and a similar meta-analysis comparing valproate and carbamazepine (Marson 2000) have failed to find convincing evidence of differences in effect between different drugs, it seems unlikely that differing preparations of the same drug are likely to have a major effect.

Agreements and disagreements with other studies or reviews

No single trial has found convincing differences between phenytoin and valproate with respect to seizure control or seizure type (Callaghan 1985; Craig 1994; Czapinski 1997a; De Silva 1996; Forsythe 1991; Heller 1995; Ramsay 1992; Rastogi 1991; Shakir 1981; Thilothammal 1996; Turnbull 1985). However, confidence intervals around estimates have been wide and equivalence cannot be inferred. Furthermore, this systematic review and meta-analysis has not found any statistically significant differences between phenytoin and valproate for any of the outcomes measures. To our knowledge, this is the only systematic review and meta-analysis which compares phenytoin and valproate monotherapy for partial onset seizures and generalised onset tonic-clonic seizures.

Authors' conclusions

Implications for practice

The results of this systematic review do not provide any conclusive evidence for or against the current practice of using valproate as a first-line treatment for individuals with generalised onset tonic-clonic seizures and phenytoin as monotherapy for individuals with partial onset seizures. Guidelines currently recommend lamotrigine as a first-line treatment for partial onset seizures (Marson 2007); the results of this review do not inform current treatment policy.

Implications for research

Finding overall differences between these standard antiepileptic drugs has proved elusive. If overall differences do exist across heterogeneous populations of individuals such as those studied here, those differences are likely to be small, and in order to be clinically useful, future comparative antiepileptic drug trials will need to be powered accordingly. It has been argued that future comparative antiepileptic drug trials be powered to establish equivalence (Jones 1996), and therefore be capable of detecting what is considered to be the smallest important clinical difference. A network meta-analysis has been published (Tudur Smith 2007), comparing all direct and indirect evidence from phenytoin, valproate and other standard and new antiepileptic drugs licensed for monotherapy, and also found no differences between phenytoin and valproate for the outcomes specified in this review. This review and the network meta-analysis will be updated as more information becomes available.

This review highlights the need for future antiepileptic drug monotherapy trials that recruit individuals with specific epilepsy syndromes to be designed and powered to detect a difference between particular antiepileptic drugs. An approach likely to reflect and inform clinical practice, as well as being statistically powerful, would be to recruit heterogeneous populations for whom epilepsy syndromes have been adequately defined, with testing for interaction between treatment and epilepsy syndrome. In view of potential problems of misclassification, syndromes will have to be well-defined, with adequate checking mechanisms to ensure that classifications are accurate, and a system to recognise uncertainty surrounding epilepsy syndromes in individuals within trials.

Clinical uncertainty about seizure and syndrome classification is often present at the time of diagnosis and initial treatment of epilepsy, and significant numbers of individuals with newly-diagnosed epilepsy cannot be classified (Bodensteiner 1988; Ottman 1993). Seizures may have been few and unwitnessed, and investigations are commonly unhelpful, but there is nevertheless no doubt that seizures have occurred and should be treated. This most commonly applies to tonic-clonic seizures that may be generalised at onset, or which may be secondarily generalised. In any trial, such unclassified individuals need to be clearly identified, because if they are not they may confound interpretation of results for well-classified individuals. We need to know how to manage those whose classification we find more difficult.

Acknowledgements

We are greatly indebted to all of the trialists who have provided individual patient data and input and review. They have shown great patience in the way our data queries were handled.
Kenneth Sommerville and Roger Deaton at Abbott Laboratories.
We acknowledge Paula Williamson for contributions to the original review

Data and analyses

Download statistical data

Comparison 1. Phenytoin versus sodium valproate
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Time to withdrawal of allocated treatment6569Hazard Ratio (Fixed, 95% CI)1.02 [0.73, 1.42]
2 Time to withdrawal of allocated treatment - stratified by epilepsy type5528Hazard Ratio (Fixed, 95% CI)1.09 [0.76, 1.55]
2.1 Generalised onset seizures (tonic-clonic only)5341Hazard Ratio (Fixed, 95% CI)0.98 [0.59, 1.64]
2.2 Partial onset seizures4187Hazard Ratio (Fixed, 95% CI)1.20 [0.74, 1.95]
3 Time to 12-month remission4514Hazard Ratio (Fixed, 95% CI)0.97 [0.77, 1.22]
4 Time to 12-month remission - stratified by epilepsy type4514Hazard Ratio (Fixed, 95% CI)0.98 [0.78, 1.23]
4.1 Generalised onset seizures (tonic-clonic only)4270Hazard Ratio (Fixed, 95% CI)1.04 [0.77, 1.40]
4.2 Partial onset seizures4244Hazard Ratio (Fixed, 95% CI)0.90 [0.63, 1.29]
5 Time to six-month remission5639Hazard Ratio (Fixed, 95% CI)0.92 [0.76, 1.12]
6 Time to six-month remission - stratified by epilepsy type5639Hazard Ratio (Fixed, 95% CI)0.95 [0.78, 1.15]
6.1 Generalised onset seizures (tonic-clonic only)5395Hazard Ratio (Fixed, 95% CI)0.92 [0.72, 1.18]
6.2 Partial onset seizures4244Hazard Ratio (Fixed, 95% CI)0.99 [0.73, 1.35]
7 Time to first seizure5639Hazard Ratio (Fixed, 95% CI)0.96 [0.78, 1.18]
8 Time to first seizure - stratified by epilepsy type5639Hazard Ratio (Fixed, 95% CI)0.93 [0.75, 1.14]
8.1 Generalised onset seizures (tonic-clonic only)5395Hazard Ratio (Fixed, 95% CI)1.03 [0.77, 1.39]
8.2 Partial onset seizures4244Hazard Ratio (Fixed, 95% CI)0.83 [0.62, 1.11]
9 Time to first seizure - epilepsy type reclassified to uncertain for generalised and age of onset > 30 years5649Hazard Ratio (Fixed, 95% CI)0.93 [0.76, 1.15]
9.1 Generalised onset seizures (tonic-clonic only)4223Hazard Ratio (Fixed, 95% CI)1.34 [0.91, 1.97]
9.2 Partial onset seizures4255Hazard Ratio (Fixed, 95% CI)0.83 [0.62, 1.11]
9.3 Uncertain seizure type4171Hazard Ratio (Fixed, 95% CI)0.74 [0.47, 1.17]
10 Time to first seizure - epilepsy type reclassified to partial for generalised and age of onset >30 years5639Hazard Ratio (Fixed, 95% CI)0.94 [0.76, 1.15]
10.1 Generalised onset seizures (tonic-clonic only)4223Hazard Ratio (Fixed, 95% CI)1.34 [0.91, 1.97]
10.2 Partial onset seizures5416Hazard Ratio (Fixed, 95% CI)0.81 [0.64, 1.04]
Analysis 1.1.

Comparison 1 Phenytoin versus sodium valproate, Outcome 1 Time to withdrawal of allocated treatment.

Analysis 1.2.

Comparison 1 Phenytoin versus sodium valproate, Outcome 2 Time to withdrawal of allocated treatment - stratified by epilepsy type.

Analysis 1.3.

Comparison 1 Phenytoin versus sodium valproate, Outcome 3 Time to 12-month remission.

Analysis 1.4.

Comparison 1 Phenytoin versus sodium valproate, Outcome 4 Time to 12-month remission - stratified by epilepsy type.

Analysis 1.5.

Comparison 1 Phenytoin versus sodium valproate, Outcome 5 Time to six-month remission.

Analysis 1.6.

Comparison 1 Phenytoin versus sodium valproate, Outcome 6 Time to six-month remission - stratified by epilepsy type.

Analysis 1.7.

Comparison 1 Phenytoin versus sodium valproate, Outcome 7 Time to first seizure.

Analysis 1.8.

Comparison 1 Phenytoin versus sodium valproate, Outcome 8 Time to first seizure - stratified by epilepsy type.

Analysis 1.9.

Comparison 1 Phenytoin versus sodium valproate, Outcome 9 Time to first seizure - epilepsy type reclassified to uncertain for generalised and age of onset > 30 years.

Analysis 1.10.

Comparison 1 Phenytoin versus sodium valproate, Outcome 10 Time to first seizure - epilepsy type reclassified to partial for generalised and age of onset >30 years.

Appendices

Appendix 1. Cochrane Epilepsy Group's Specialized Register search strategy

#1 phenytoin or Epanutin or Phenytek or Dilantin or Eptoin or Diphenin or Dipheninum or Diphenylhydantoin

#2 Depakene or Depacon or Depakine or Valparin or Stavzor or Epilim or Epiject or Episenta or Epival or Valpro* or Orlept or Orfiril or Selenica or Convulex or Depakote

#3 #1 AND #2

#4 (adjunct* or "add-on" or "add on") not monotherap*

#5 #3 NOT #4

Appendix 2. CENTRAL search strategy

#1         MeSH descriptor: [Phenytoin] explode all trees

#2         Epanutin or Phenytek or Dilantin or Eptoin or Diphenin or Dipheninum or Diphenylhydantoin:ti,ab,kw  (Word variations have been searched)

#3         #1 or #2

#4         MeSH descriptor: [Valproic Acid] explode all trees

#5         Depakene or Depacon or Depakine or Valparin or Stavzor or Epilim or Epiject or Episenta or Epival or Valpro* or Orlept or Orfiril or Selenica or Convulex or Depakote:ti,ab,kw  (Word variations have been searched)

#6         #4 or #5

#7         #3 and #6

#8         (adjunct* or "add-on" or "add on") not monotherap*:ti,ab,kw  (Word variations have been searched)

#9         #7 not #8

#10       (epilep* or seizure* or convuls*):ti,ab,kw  (Word variations have been searched)

#11       MeSH descriptor: [Epilepsy] explode all trees

#12       MeSH descriptor: [Seizures] explode all trees

#13       (#10 or #11 or #12) in Trials

#14       #9 and #13

Appendix 3. MEDLINE search strategy

The following search is based on the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE (Lefebvre 2011).

1. exp phenytoin/ or (Epanutin or Phenytek or Dilantin or Eptoin or Diphenin or Dipheninum or Diphenylhydantoin).mp.

2. exp Valproic Acid/ or (Depakene or Depacon or Depakine or Valparin or Stavzor or Epilim or Epiject or Episenta or Epival or Valpro$ or Orlept or Orfiril or Selenica or Convulex or Depakote).mp.

3. ((adjunct$ or "add-on" or "add on") not monotherap$).mp.

4. (1 and 2) not 3

5. (randomized controlled trial or controlled clinical trial).pt. or (randomized or placebo or randomly).ab.

6. clinical trials as topic.sh.

7. trial.ti.

8. 5 or 6 or 7

9. exp animals/ not humans.sh.

10. 8 not 9

11. exp Epilepsy/

12. exp Seizures/

13. (epilep$ or seizure$ or convuls$).tw.

14. 11 or 12 or 13

15. exp Pre-Eclampsia/ or exp Eclampsia/

16. 14 not 15

17. 4 and 10 and 16

Earlier versions of this review used the following search, based on the previous Cochrane Highly Sensitive Search Strategy for MEDLINE as set out in Appendix 5b of the Cochrane Handbook for Systematic Reviews of Interventions (version 4.2.4, updated March 2005) (Higgins 2011).

1. randomized controlled trial.pt.

2. controlled clinical trial.pt.

3. exp Randomized Controlled Trials/

4. exp Random Allocation/

5. exp Double-Blind Method/

6. exp Single-Blind Method/

7. clinical trial.pt.

8. Clinical Trial/

9. (clin$ adj trial$).ab,ti.

10. ((singl$ or doubl$ or trebl$ or tripl$) adj (blind$ or mask$)).ab,ti.

11. exp PLACEBOS/

12. placebo$.ab,ti.

13. random$.ab,ti.

14. exp Research Design/

15. or/1-14

16. (animals not humans).sh.

17. 15 not 16

18. phenytoin/ or (phenytoin or diphenylhydantoin).tw.

19. valproic acid/ or valpro$.tw.

20. exp epilepsy/ or epilep$.tw.

21. exp seizures/ or seizure$.tw.

22. convulsion$.tw.

23. 18 and 19

24. 20 or 21 or 22

25. 23 and 24

26. 17 and 25

Appendix 4. SCOPUS search strategy

(((TITLE(phenytoin or Epanutin or Phenytek or Dilantin or Eptoin or Diphenin or Dipheninum or Diphenylhydantoin) or ABS(phenytoin or Epanutin or Phenytek or Dilantin or Eptoin or Diphenin or Dipheninum or Diphenylhydantoin)) and (TITLE(Depakene or Depacon or Depakine or Valparin or Stavzor or Epilim or Epiject or Episenta or Epival or Valpro* or Orlept or Orfiril or Selenica or Convulex or Depakote) or ABS(Depakene or Depacon or Depakine or Valparin or Stavzor or Epilim or Epiject or Episenta or Epival or Valpro* or Orlept or Orfiril or Selenica or Convulex or Depakote))) and not (TITLE-ABS-KEY((adjunct* OR "add-on" OR "add on") AND NOT monotherap*))) and (TITLE((randomiz* OR randomis* OR controlled OR placebo OR blind* OR unblind* OR "parallel-group" OR "parallel group" OR crossover OR cross-over OR "cross over" OR cluster OR "head to head" OR "head-to-head") PRE/2 (trial OR method OR procedure OR study)) OR ABS((randomiz* OR randomis* OR controlled OR placebo OR blind* OR unblind* OR "parallel-group" OR "parallel group" OR crossover OR cross-over OR "cross over" OR cluster OR "head to head" OR "head-to-head") PRE/2 (trial OR method OR procedure OR study))) and ((TITLE-ABS-KEY(epilep* OR "infantile spasm" OR seizure OR convuls* OR (syndrome W/2 (aicardi OR angelman OR doose OR dravet OR janz OR jeavons OR "landau kleffner" OR "lennox gastaut" OR ohtahara OR panayiotopoulos OR rasmussen OR rett OR "sturge weber" OR tassinari OR "unverricht lundborg" OR west)) OR "ring chromosome 20" OR "R20" OR "myoclonic encephalopathy" OR "pyridoxine dependency") AND NOT (TITLE(*eclampsia) OR INDEXTERMS(*eclampsia))) OR (TITLE-ABS-KEY(lafora* W/4 (disease OR epilep*)) AND NOT (TITLE(dog OR canine) OR INDEXTERMS(dog OR canine))))

What's new

DateEventDescription
13 August 2013New citation required but conclusions have not changedConclusions unchanged.
21 February 2013New search has been performedSearches updated February 2013. Analyses and text updated. 'Risk of bias' assessments and 'Summary of findings' table added.

History

Protocol first published: Issue 3, 1999
Review first published: Issue 4, 2001

DateEventDescription
23 September 2008AmendedConverted to new review format.
27 July 2007New search has been performedWe re-ran our searches on 27 July 2007; one new study has been identified and added to the 'studies awaiting assessment' section. It will be assessed for inclusion in the review at a later date.

Contributions of authors

SJ Nolan assessed studies for inclusion in the review update, assessed risk of bias in all included studies, performed analyses in SAS version 9.2, Stata version 11.2 and Metaview, added survival plots and a 'Summary of findings' table and updated the text of the review under the supervision of C Tudur Smith and AG Marson.

C Tudur Smith was the lead investigator on the original review, assessed eligibility and methodological quality of original individual studies, organised and cleaned the individual patient data sets, performed data validation checks and statistical analyses and co-wrote the original review.

AG Marson obtained individual patient data from trial investigators, provided guidance with the clinical interpretation of results, assessed eligibility and methodological quality of individual studies and co-wrote the original review.

J Pulman independently assessed risk of bias in all included studies

Declarations of interest

None known.

Sources of support

Internal sources

  • University of Liverpool, UK.

  • Walton Centre for Neurology and Neurosurgery, UK.

External sources

  • Medical Research Council, UK.

  • NHS R&D, UK.

Notes

The protocol for this review was published with Catrin Tudur as the contact review author. Catrin is now known as Catrin Tudur Smith.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Callaghan 1985

Methods

Parallel study design, outpatient setting

Study conducted in Eire (Republic of Ireland)

Randomisation based on two Latin squares and the preference of drug for the participant.

An independent person selected “drug of first preference” from randomisation list.

Participants

Adults and children with a minimum of 2 untreated generalised or partial seizures in the 6 months preceding the trial.

Number randomised: PHT = 58; SV = 64

48 participants (39%) with partial epilepsy. 67 (55%) men.

Age range: 5 - 71. Duration of treatment (range in months):3 - 48

Interventions

Monotherapy with PHT or SV.

Mean daily dose achieved: PHT: 5.4 mg/kg; SV: 15.6 mg/kg.

OutcomesSeizure control:
excellent (complete freedom of seizures);
good (> 50% reduction in seizure frequency);
poor (< 50% reduction in seizure frequency).
NotesOutcomes chosen for this review were not reported. IPD not available.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomisation based on 2 Latin Squares without stratification. The first, second and third preference of drug for the participant appears to have been taken into account in the process. Unclear if assignment was completely random
Allocation concealment (selection bias)High riskAn independent person (department secretary) selected the “drug of first preference” from randomisation list on a sequential basis. Allocation not adequately concealed
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo information provided
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNo information provided
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttirition rates reported. ITT approach taken, all randomised participants analysed
Selective reporting (reporting bias)Low riskPrimary outcomes (seizure control) and secondary outcomes (side effects) reported sufficiently. No protocol available, outcomes for this review not reported
Other biasLow riskNo other bias detected

Craig 1994

Methods

Parallel study design

Study conducted in the UK

Participants randomised using computerised stratified minimisation program by age group, sex and seizure type.

Allocation was pharmacy controlled.

The main investigator performing cognitive testing was blinded to allocation. Participants and personnel unblinded

Participants

Participants over 60 years of age with newly onset seizures (1 or more generalised tonic-clonic seizures or 2 or more partial seizures).

Number randomised: PHT:81, SV:85.

80 participants (48%) with partial epilepsy, 71 (44%) men.

Mean age (range): 78 (61 - 95 years). Range of follow-up: 1 - 20 months

Interventions

Monotherapy with PHT or SV.

Starting doses: PHT: 200 mg/day, SV: 400 mg/day.

Median daily dose achieved: PHT 247 mg (range 175 - 275); SV: 688 mg (range 400 - 1000)

Outcomes

Psychological tests (cognitive function, anxiety and depression)

Adverse event frequency

Seizure control

NotesTrial paper reports on a subset of 38 participants. Full individual patient data set provided and used for this review includes all 166 participants randomised in the trial. IPD provided for 3/4 outcomes of this review ('withdrawal from allocated treatment' not available).
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskComputerised stratified minimisation programme, stratified for age group, gender and seizure type
Allocation concealment (selection bias)Low riskPharmacy-controlled allocation, prescription disclosed to general practitioner and consultant
Blinding of participants and personnel (performance bias)
All outcomes
High riskParticipants and personnel unblinded
Blinding of outcome assessment (detection bias)
All outcomes
Low riskThe main investigator performing cognitive testing was blinded to allocation
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttrition rates reported. ITT analysis undertaken with all randomised participants from IPD (see footnote 2)
Selective reporting (reporting bias)Low riskAll outcome measures reported in published report or provided in IPD (see footnote 2)
Other biasLow riskNo other bias detected

Czapinski 1997a

Methods

36-month randomised comparative trial

Parallel study design

Study conducted in Poland

Method of generation of random list and allocation concealment not stated.

Participants

Adults with newly-diagnosed epilepsy.

Number randomised: PHT: 30, SV: 30

100% partial epilepsy, Age range: 18 to 40 years

Percentage men and range of follow-up not mentioned.

Interventions

Monotherapy with PHT or SV.

Starting doses: PHT: 200 mg/day, SV: 600 mg/day. Dose achieved not stated.

OutcomesProportion achieving 24-month remission at 3 years
exclusions after randomisation due to adverse events or no efficacy.
NotesAbstract only. Outcomes chosen for this review were not reported. IPD pledged but not received.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskTrial "randomised" but no further information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo information provided
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNo information provided
Incomplete outcome data (attrition bias)
All outcomes
Unclear risk"Exclusion rates" (interpreted as withdrawal rates) reported for all treatment groups, no further information provided
Selective reporting (reporting bias)Unclear riskNo protocol available and trial reported only in abstract form, outcomes for this review not available
Other biasUnclear riskInsufficient detail provided in abstract to allow judgement

De Silva 1996

Methods

Parallel study design, outpatient setting

Study conducted at two centres in the UK

Random list generated using random permuted blocks.

Allocation concealed using sealed opaque envelopes

Unblinded

Participants

Children with newly-diagnosed epilepsy (2 or more untreated partial or generalised tonic clonic seizures in the 12 months preceding the trial).

Number randomised: PHT: 54; SV: 49.

55 children (53%) with partial epilepsy. 52 (50%) boys.

Mean age (range): 10 (3 - 16) years. Range of follow-up (months): 3 - 88.

Interventions

Monotherapy with PHT or SV.

Median daily dose achieved: PHT: 175 mg/day, SV: 600 mg/day.

OutcomesTime to first seizure recurrence after start of therapy.
Time to 12-month remission from all seizures.
Adverse events and withdrawals due to adverse events
NotesIPD provided for all outcomes of this review
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandomisation list generated using permuted blocks of size 8 or 16 with stratification for centre, seizure type and presence of neurological signs
Allocation concealment (selection bias)Low riskAllocation concealed via 4 batches of sealed opaque envelopes
Blinding of participants and personnel (performance bias)
All outcomes
High riskUnblinded, authors state masking of treatment would not be “practicable or ethical” and would “undermine compliance.”
Blinding of outcome assessment (detection bias)
All outcomes
High riskUnblinded, authors state masking of treatment would not be “practicable or ethical” and would “undermine compliance.”
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttrition rates reported, all randomised participants analysed from IPD provided (see footnote 2)
Selective reporting (reporting bias)Low riskAll outcomes reported or calculated with IPD provided (see footnote 2)
Other biasLow riskNo other bias detected

Forsythe 1991

Methods

Parallel study design, outpatient setting

Study conducted in the UK

Patients randomly allocated using quota allocation allowing for gender, age, seizure type and current treatment.

Outcome assessors were single-blinded for cognitive testing

Participants

Children with at least 3 newly diagnosed generalised or partial seizures within a period of 6 months.

Number randomised: PHT: 20; SV: 21.

No information on epilepsy type, gender or range of follow-up.

Age range: 5 - 14 years. Trial duration: 12 months

Interventions

Monotherapy with PHT or SV.

Mean dose achieved: PHT: 6.1 mg/day, SV: 25.3 mg/day.

OutcomesCognitive assessments.
Summary of withdrawals from randomised drug.
NotesOutcomes chosen for this review were not reported. IPD not available, but could be constructed from the publication for the outcome 'Time on allocated drug' (without stratification by seizure type)
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)High riskQuota allocation by gender, age , seizure type and current treatment is an inadequate randomisation method
Allocation concealment (selection bias)Unclear riskNo information provided
Blinding of participants and personnel (performance bias)
All outcomes
High riskPersonnel and participants (and parents) unblinded
Blinding of outcome assessment (detection bias)
All outcomes
Low riskOutcome assessors single-blinded for cognitive testing
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttrition rates reported, results reported and analysed for all participants randomised and all who completed various stages of follow-up 
Selective reporting (reporting bias)Unclear riskCognitive outcomes described in methods section well-reported in results section. Adverse events reported, no seizure outcomes reported and outcomes chosen for this review not reported. No protocol available so unclear if seizure outcomes were planned a priori.
Other biasLow riskNo other bias detected.

Heller 1995

Methods

Parallel study design, outpatient setting

Study conducted at two centres in the UK

Random list generated using random permuted blocks.

Allocation concealed using sealed opaque envelopes

Unblinded

Participants

Adults with newly-diagnosed epilepsy (2 or more untreated partial or generalised tonic clonic seizures in the 12 months preceding the trial).

Number randomised: PHT: 63; SV: 61.

53 participants (43%) with partial epilepsy. 62 (48%) men.

Mean age (range): 33 (14 - 72) years

Range of follow-up (months): 1 - 91.

Interventions

Monotherapy with PHT or SV.

Median daily dose achieved: PHT: 300 mg/day, SV: 800 mg/day.

OutcomesTime to first seizure recurrence after start of therapy.
Time to 12-month remission from all seizures.
Adverse events and withdrawal due to adverse events
NotesIPD provided for all outcomes of this review
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandomisation list generated using permuted blocks of size 8 or 16 with stratification for centre, seizure type and presence of neurological signs
Allocation concealment (selection bias)Low riskAllocation concealed via 4 batches of concealed opaque envelopes
Blinding of participants and personnel (performance bias)
All outcomes
High riskUnblinded, authors state masking of treatment would not be “practical” and would have “introduced bias due to a very large drop-out rate.”
Blinding of outcome assessment (detection bias)
All outcomes
High riskUnblinded, authors state masking of treatment would not be “practical” and would have “introduced bias due to a very large drop-out rate.”
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttrition rates reported, all randomised participants analyses from IPD provided (see footnote 2)
Selective reporting (reporting bias)Low riskAll outcomes reported or calculated with IPD provided (see footnote 2)
Other biasLow riskNo other bias detected

Ramsay 1992

Methods

Parallel trial

Study conducted at 16 centres in the United States

Participants assigned via randomisation tables within each centre in a 2:1 ratio (SV:PHT).

Method of allocation concealment not stated

Unblinded

Participants

Participants with at least 2 newly-diagnosed and previously untreated primary generalised tonic clonic seizures within 14 days of starting the trial.

Number randomised: PHT:50, SV: 86

0% participants with partial epilepsy, 73 (54%) men.

Mean age (range): 21 (3 - 64 years). Participants followed up for up to 6 months

Interventions

Monotherapy with PHT or SV.

Starting doses PHT: 3 - 5 mg/kg/day, SV: 10 - 15 mg/kg/day, doses gradually increased.

Doses achieved not stated.

Outcomes

Time to first generalised tonic clonic seizure.

6-month seizure recurrence rates.

Adverse events

NotesIPD provided for 3/4 outcomes of this review (maximum follow-up 6 months, therefore trial cannot contribute to outcome 'Time to 12-month remission')
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskParticipants randomised on a 2:1 ratio SV:PHT using randomisation tables in each centre (information provided by trial author)
Allocation concealment (selection bias)Unclear riskNo information provided
Blinding of participants and personnel (performance bias)
All outcomes
High riskOpen-label trial; authors state that differences in adverse events of PHT and SV would "quickly unblind" the trial anyway
Blinding of outcome assessment (detection bias)
All outcomes
High riskOpen-label trial, authors state that differences in adverse events of PHT and SV would "quickly unblind" the trial anyway
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttrition rates reported, all randomised participants analysed from IPD provided (see footnote 2)
Selective reporting (reporting bias)Low riskAll outcomes reported or calculated with IPD provided (see footnote 2)
Other biasLow riskNo other bias detected

Rastogi 1991

Methods

Parallel study design, outpatient setting

Study conducted in Meerut, India

No information provided on method of generation of random list, allocation concealment or blinding.

Participants

Participants with at least 2 partial or generalised tonic-clonic seizures per month

Unclear if participants were newly diagnosed

Number randomised: PHT: 45; SV: 49

27 participants (29%) partial epilepsy, 70 (74%) men.

Age range: PHT: 12 - 42 years; SV: 8 - 52 years.

Participants were evaluated after 4, 12 and 24 weeks of treatment.

No information on range of follow-up.

Interventions

Monotherapy with PHT or SV.

Average daily dose achieved: PHT: 5.6 mg/kg/day, SV: 18.8 mg/kg/day.

Outcomes

Reduction in frequency of seizures:
excellent (100% reduction);
good (75 - 99% reduction);
fair (50 - 74% reduction);
poor (< 50% reduction).
Adverse effects.

Seizure control

NotesOutcomes chosen for this review were not reported. IPD not available.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskParticipants "randomly allocated irrespective of seizure type," no further information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo information provided
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNo information provided
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskFrequency of seizures reported for all randomised participants, no information provided on withdrawal rates/attrition rates etc
Selective reporting (reporting bias)Low risk

Frequency of seizures during treatment well reported, most common adverse events reported.

No protocol available to compare with a priori analysis plan, outcomes for this review not reported

Other biasLow riskNo other bias detected

Shakir 1981

Methods

Parallel study design, outpatient setting

Study conducted in two centres (Glasgow, Scotland and Wellington, New Zealand)

Participants allocated using telephone randomisation within the two centres (– information provided by trial author).

No information provided on method of allocation concealment or blinding

Participants

21 (64%) of participants previously untreated, 12 (36%) of participants continued to have seizures on previous drug therapies.

Original treatments gradually withdrawn before PHT or SV treatment introduced.

Number randomised: PHT 15, SV: 18.

19 participants (58%) with partial epilepsy, 12 (36%) men.

Mean age (range): 23 (7 - 55 years). Mean follow-up (range): 30 (9 - 48 months).

Interventions

Monotherapy with PHT or SV.

Starting doses: PHT: < 12 years 150 mg/day, older participants: 300 mg/day

SV: < 12 years 300 - 400 mg/day, older participants: 800 - 1200 mg/day. Doses achieved not stated.

OutcomesSeizures during treatment.
Adverse events.
Notes

Outcomes chosen for this review were not reported.

IPD not available but could be constructed from the publication for the outcome 'Time to treatment withdrawal.'

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskParticipants "randomly divided", using telephone randomisation (information provided by trial author)
Allocation concealment (selection bias)Unclear riskNo information provided
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo information provided
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNo information provided
Incomplete outcome data (attrition bias)
All outcomes
Low riskResults reported for all randomised participants, time on treatment reported for all randomised participants. No losses to follow-up reported
Selective reporting (reporting bias)Low riskNo protocol available, outcomes chosen for this review not reported. Seizure outcomes and adverse events well reported
Other biasLow riskNo other bias detected

Thilothammal 1996

Methods

Parallel study design, outpatient setting

Study conducted in Madras (Chennai), India

Random list generated using computer-generated random numbers.

Method of concealment not mentioned.

Double-blind achieved by providing additional placebo tablets.               

Participants

Children with more than 1 previously untreated generalised tonic clonic (afebrile) seizure.

Number randomised: PHT: 52, SV: 48.

0% partial epilepsy. 52 (52%) men. Age range: 4 - 12 years.

Range of follow-up (months): 22 - 36.

Interventions

Monotherapy with PHT or SV.

Starting doses: PHT: 5 - 8 mg/kg/day, SV: 15 - 50 mg/kg/day.

Dose achieved not stated.

OutcomesProportion with recurrence of seizures.
Adverse events
NotesOutcomes chosen for this review were not reported. IPD not available.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskParticipants randomised via a computer-generated list of random numbers
Allocation concealment (selection bias)Unclear riskNo information provided
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskDouble–blinded using additional placebo tablets, unclear who was blinded
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskDouble–blinded using additional placebo tablets, unclear who was blinded
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttrition rates reported, all randomised participants analysed
Selective reporting (reporting bias)Low riskNo protocol available, outcomes chosen for this review not reported
Other biasLow riskNo other bias detected

Turnbull 1985

  1. a

    1 Abbreviations:
    AED: antiepileptic drug
    IPD: individual patient data
    ITT: Intention-to-treat
    PHT: phenytoin
    SV: sodium valproate

    2 For studies which provided IPD (Craig 1994;De Silva 1996; Heller 1995; Ramsay 1992; Turnbull 1985) attrition and reporting bias are reduced as attrition rates and unpublished outcome data are requested.

    3 See Figure 2; and Figure 3 for Risk of Bias graphically.

Methods

Parallel study design, outpatient setting

Study conducted in the UK

Participants allocated to treatment stratified by age group, gender and seizure type.

No information provided on method of generation of random list, allocation concealment or blinding.

Participants

Participants with 2 or more partial or generalised tonic-clonic seizure in the past 3 years.

Participants were previously untreated but started on AED treatment within 3 months of their most recent seizure.

Number randomised: PHT:70, SV:70.

63 participants (45%) with partial onset seizures, 73 (52%) men.

Mean age (range): 35 (14 - 70 years). Range of follow-up: 24 - 48 months

Interventions

Monotherapy with PHT or SV.

Starting doses: PHT 300 mg/day, SV 600 mg/day. Dose achieved not stated 

Outcomes

Time to 2-year remission

Time to first seizure

Adverse events

NotesIPD provided for all outcomes included in this review.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskParticipants randomised with stratification for age group, gender and seizure type. Method of randomisation not stated
Allocation concealment (selection bias)Unclear riskNo information provided
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo information provided
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskNo information provided
Incomplete outcome data (attrition bias)
All outcomes
Low riskAttrition rates reported, ITT approach, all randomised participants analysed from IPD provided (see footnote 2)
Selective reporting (reporting bias)Low riskAll outcomes reported or calculated with IPD provided (see footnote 2)
Other biasUnclear riskNo other bias detected

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion
  1. a

    AED: antiepileptic drug

Berg 1993Reports the same trial as Forsythe 1991, but more relevant information given in the Forsythe publication.
Callaghan 1981Abstract only. Preliminary results of the trial reported in Callaghan 1985
Callaghan 1983Abstract only. Preliminary results of the trial reported in Callaghan 1985
Callaghan 1984Preliminary results of the trial reported in Callaghan 1985
Craig 1993Abstract only. Prelimiary results of the trial reported in Craig 1994.
Czapinski 1997bReports the same abstract as Czapinski 1997a
Czapinski 1997cReports the same abstract as Czapinski 1997a
Goggin 1984Abstract only. Preliminary results of the trial reported in Callaghan 1985
Goggin 1986Reports the same trial as Callaghan 1985, but more relevant information given in the Callaghan publication.
Jannuzzi 2000No randomised comparison of phenytoin and valproate (participants randomised to a dose adjustment method rather than to a treatment)
Kaminow 2003No randomised comparison of phenytoin and valproate (study of lamotrigine vs 'standard' AED treatment)
Sabers 1995Not fully randomised: "The treatment was chosen at random unless the individual diagnoses required a specific drug"
Schmidt 2007No randomised comparison of phenytoin and valproate (post hoc analysis of 5 studies of oxcarbazepine vs another AED)
Shakir 1980Reports the same trial as Shakir 1981. There are some differences between the results in the 2 publications. The reason for this could not be established.
Tallis 1994aAbstract only. Reports the same trial as Craig 1994.
Tallis 1994bAbstract only. Reports the same trial as Craig 1994.
Turnbull 1982Preliminary results of the trial reported in Turnbull 1985
Wilder 1983Preliminary results of the trial reported in Turnbull 1985.
Zeng 2010Not randomised

Ancillary